release 0.4.25 (#969)

* add teh system role into messages array

* ci: trigger workflows

* dont normalize for anthropic

---------

Co-authored-by: Spherrrical <malikmusa1323@gmail.com>

.github/workflows/ci.yml

@@ -107,6 +107,56 @@ jobs:
         if: always()
         run: planoai down || true
 
+      # ── Zero-config path: `planoai up` with no args, no plano.yaml in cwd.
+      # Exercises the synthesize_default_config branch in cli/planoai/main.py
+      # which is otherwise never hit by the smoke test above.
+      #
+      # Pre-seed ~/.plano/ from the freshly-built artifacts so the CLI's
+      # cached-download path hits in step (2) of ensure_wasm_plugins /
+      # ensure_brightstaff_binary. Without this, running from outside the
+      # repo means find_repo_root() returns None, the local-build short-
+      # circuit is skipped, and the CLI tries to download from a GitHub
+      # release that does not yet exist for the in-flight version on
+      # release-bump PRs (e.g. 0.4.25 before publish-binaries has run).
+      - name: Seed ~/.plano cache for zero-config test
+        run: |
+          VERSION=$(sed -nE 's/^__version__ = "(.*)"$/\1/p' cli/planoai/__init__.py)
+          mkdir -p ~/.plano/plugins ~/.plano/bin
+          cp crates/target/wasm32-wasip1/release/prompt_gateway.wasm ~/.plano/plugins/
+          cp crates/target/wasm32-wasip1/release/llm_gateway.wasm ~/.plano/plugins/
+          cp crates/target/release/brightstaff ~/.plano/bin/
+          chmod +x ~/.plano/bin/brightstaff
+          echo "$VERSION" > ~/.plano/plugins/wasm.version
+          echo "$VERSION" > ~/.plano/bin/brightstaff.version
+
+      - name: Zero-config smoke test
+        env:
+          OPENAI_API_KEY: test-key-not-used
+        run: |
+          empty_dir="$(mktemp -d)"
+          cd "$empty_dir"
+          test ! -f plano.yaml
+          planoai up
+          test -f "$HOME/.plano/default_config.yaml"
+
+      - name: Zero-config health check
+        run: |
+          for i in $(seq 1 30); do
+            if curl -sf http://localhost:12000/healthz > /dev/null 2>&1; then
+              echo "Zero-config health check passed"
+              exit 0
+            fi
+            sleep 1
+          done
+          echo "Zero-config health check failed after 30s"
+          cat ~/.plano/run/logs/envoy.log || true
+          cat ~/.plano/run/logs/brightstaff.log || true
+          exit 1
+
+      - name: Stop plano (zero-config)
+        if: always()
+        run: planoai down || true
+
   # ──────────────────────────────────────────────
   # Single Docker build — shared by all downstream jobs
   # ──────────────────────────────────────────────
@@ -133,13 +183,13 @@ jobs:
           load: true
           tags: |
             ${{ env.PLANO_DOCKER_IMAGE }}
-            ${{ env.DOCKER_IMAGE }}:0.4.22
+            ${{ env.DOCKER_IMAGE }}:0.4.25
             ${{ env.DOCKER_IMAGE }}:latest
           cache-from: type=gha
           cache-to: type=gha,mode=max
 
       - name: Save image as artifact
-        run: docker save ${{ env.PLANO_DOCKER_IMAGE }} ${{ env.DOCKER_IMAGE }}:0.4.22 ${{ env.DOCKER_IMAGE }}:latest -o /tmp/plano-image.tar
+        run: docker save ${{ env.PLANO_DOCKER_IMAGE }} ${{ env.DOCKER_IMAGE }}:0.4.25 ${{ env.DOCKER_IMAGE }}:latest -o /tmp/plano-image.tar
 
       - name: Upload image artifact
         uses: actions/upload-artifact@v6

.github/workflows/update-providers.yml

@@ -0,0 +1,124 @@
+name: Update provider_models.yaml
+
+on:
+  repository_dispatch:
+    types: [update-providers]
+  workflow_dispatch:
+
+permissions:
+  contents: write
+  pull-requests: write
+
+jobs:
+  update-providers:
+    runs-on: ubuntu-latest
+    env:
+      RESPONSE_URL: ${{ github.event.client_payload.response_url }}
+      SLACK_USER_ID: ${{ github.event.client_payload.user_id }}
+      SLACK_USER_NAME: ${{ github.event.client_payload.user_name }}
+    steps:
+      - name: Checkout main
+        uses: actions/checkout@v6
+        with:
+          ref: main
+
+      - name: Install Rust toolchain
+        uses: dtolnay/rust-toolchain@stable
+
+      - name: Configure AWS credentials
+        uses: aws-actions/configure-aws-credentials@v4
+        with:
+          aws-access-key-id: ${{ secrets.AWS_ACCESS_KEY_ID }}
+          aws-secret-access-key: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
+          aws-region: ${{ secrets.AWS_REGION }}
+
+      - name: Cache cargo build
+        uses: actions/cache@v4
+        with:
+          path: |
+            ~/.cargo/registry
+            ~/.cargo/git
+            crates/target
+          key: cargo-fetch-models-${{ hashFiles('crates/**/Cargo.lock', 'crates/**/Cargo.toml') }}
+          restore-keys: cargo-fetch-models-
+
+      - name: Run fetch_models
+        working-directory: crates/hermesllm
+        env:
+          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
+          ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
+          MISTRAL_API_KEY: ${{ secrets.MISTRAL_API_KEY }}
+          DEEPSEEK_API_KEY: ${{ secrets.DEEPSEEK_API_KEY }}
+          GROK_API_KEY: ${{ secrets.GROK_API_KEY }}
+          DASHSCOPE_API_KEY: ${{ secrets.DASHSCOPE_API_KEY }}
+          MOONSHOT_API_KEY: ${{ secrets.MOONSHOT_API_KEY }}
+          ZHIPU_API_KEY: ${{ secrets.ZHIPU_API_KEY }}
+          MIMO_API_KEY: ${{ secrets.MIMO_API_KEY }}
+          GOOGLE_API_KEY: ${{ secrets.GOOGLE_API_KEY }}
+          OPENROUTER_API_KEY: ${{ secrets.OPENROUTER_API_KEY }}
+          AI_GATEWAY_API_KEY: ${{ secrets.AI_GATEWAY_API_KEY }}
+        run: cargo run --bin fetch_models --features model-fetch
+
+      - name: Create pull request
+        id: cpr
+        uses: peter-evans/create-pull-request@v7
+        with:
+          branch: bot/update-providers-${{ github.run_id }}
+          base: main
+          commit-message: "chore: refresh provider_models.yaml"
+          title: "chore: refresh provider_models.yaml"
+          body: |
+            Automated refresh of `crates/hermesllm/src/bin/provider_models.yaml`
+            via `fetch_models`.
+
+            Requested by ${{ env.SLACK_USER_NAME && format('@{0}', env.SLACK_USER_NAME) || 'workflow_dispatch' }}${{ env.SLACK_USER_ID && format(' (Slack `{0}`)', env.SLACK_USER_ID) || '' }}.
+
+            Workflow run: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}
+          labels: automated, provider-models
+          add-paths: crates/hermesllm/src/bin/provider_models.yaml
+
+      - name: Notify Slack (success)
+        if: success() && env.RESPONSE_URL != ''
+        env:
+          PR_URL: ${{ steps.cpr.outputs.pull-request-url }}
+          PR_NUMBER: ${{ steps.cpr.outputs.pull-request-number }}
+          PR_OP: ${{ steps.cpr.outputs.pull-request-operation }}
+        run: |
+          if [ -z "$PR_URL" ]; then
+            TEXT=":information_source: No provider model changes detected \u2014 nothing to PR."
+            BLOCKS=$(jq -nc --arg text "$TEXT" '{response_type:"ephemeral", replace_original:true, text:$text, blocks:[{type:"section", text:{type:"mrkdwn", text:$text}}]}')
+          else
+            TEXT=":white_check_mark: provider_models.yaml PR ready: $PR_URL"
+            BLOCKS=$(jq -nc \
+              --arg pr "$PR_URL" \
+              --arg num "$PR_NUMBER" \
+              --arg op "$PR_OP" \
+              '{
+                response_type:"ephemeral",
+                replace_original:true,
+                text:(":white_check_mark: provider_models.yaml PR #" + $num + " " + $op + ": " + $pr),
+                blocks:[
+                  {type:"section", text:{type:"mrkdwn", text:(":white_check_mark: *provider_models.yaml* PR <" + $pr + "|#" + $num + "> " + $op + ".")}},
+                  {type:"actions", elements:[{type:"button", text:{type:"plain_text", text:"Open PR"}, url:$pr}]}
+                ]
+              }')
+          fi
+          curl -sS -X POST -H 'Content-Type: application/json' -d "$BLOCKS" "$RESPONSE_URL"
+
+      - name: Notify Slack (failure)
+        if: failure() && env.RESPONSE_URL != ''
+        run: |
+          RUN_URL="${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}"
+          TEXT=":x: provider_models.yaml update failed. Logs: $RUN_URL"
+          jq -nc \
+            --arg text "$TEXT" \
+            --arg run "$RUN_URL" \
+            '{
+              response_type:"ephemeral",
+              replace_original:true,
+              text:$text,
+              blocks:[
+                {type:"section", text:{type:"mrkdwn", text:(":x: *provider_models.yaml update failed.*")}},
+                {type:"actions", elements:[{type:"button", text:{type:"plain_text", text:"View logs"}, url:$run}]}
+              ]
+            }' | curl -sS -X POST -H 'Content-Type: application/json' -d @- "$RESPONSE_URL"

CLAUDE.md

@@ -49,7 +49,7 @@ Client → Envoy (prompt_gateway.wasm → llm_gateway.wasm) → Agents/LLM Provi
 
 ### Python CLI (cli/planoai/)
 
-Entry point: `main.py`. Built with `rich-click`. Commands: `up`, `down`, `build`, `logs`, `trace`, `init`, `cli_agent`, `generate_prompt_targets`.
+Entry point: `main.py`. Built with `rich-click`. Commands: `up`, `down`, `build`, `logs`, `trace`, `init`, `cli_agent`.
 
 ### Config (config/)
 

README.md

@@ -32,7 +32,7 @@ Plano solves this by moving core delivery concerns into a unified, out-of-proces
 Plano pulls rote plumbing out of your framework so you can stay focused on what matters most: the core product logic of your agentic applications. Plano is backed by [industry-leading LLM research](https://planoai.dev/research) and built on [Envoy](https://envoyproxy.io) by its core contributors, who built critical infrastructure at scale for modern worklaods.
 
 **High-Level Network Sequence Diagram**:
-![high-level network plano arcitecture for Plano](docs/source/_static/img/plano_network_diagram_high_level.png)
+![high-level network plano architecture for Plano](docs/source/_static/img/plano_network_diagram_high_level.png)
 
 **Jump to our [docs](https://docs.planoai.dev)** to learn how you can use Plano to improve the speed, safety and obervability of your agentic applications.
 
@@ -156,7 +156,7 @@ curl http://localhost:8001/v1/chat/completions \
 
 Every request is traced end-to-end with OpenTelemetry - no instrumentation code needed.
 
-![Atomatic Tracing](docs/source/_static/img/demo_tracing.png)
+![Automatic Tracing](docs/source/_static/img/demo_tracing.png)
 
 ### What You Didn't Have to Build
 
@@ -183,7 +183,6 @@ Ready to try Plano? Check out our comprehensive documentation:
 - **[LLM Routing](https://docs.planoai.dev/guides/llm_router.html)** - Route by model name, alias, or intelligent preferences
 - **[Agent Orchestration](https://docs.planoai.dev/guides/orchestration.html)** - Build multi-agent workflows
 - **[Filter Chains](https://docs.planoai.dev/concepts/filter_chain.html)** - Add guardrails, moderation, and memory hooks
-- **[Prompt Targets](https://docs.planoai.dev/concepts/prompt_target.html)** - Turn prompts into deterministic API calls
 - **[Observability](https://docs.planoai.dev/guides/observability/observability.html)** - Traces, metrics, and logs
 
 ## Contribution

apps/www/src/components/Hero.tsx

@@ -24,7 +24,7 @@ export function Hero() {
             >
               <div className="inline-flex flex-wrap items-center gap-1.5 sm:gap-2 px-3 sm:px-4 py-1 rounded-full bg-[rgba(185,191,255,0.4)] border border-[var(--secondary)] shadow backdrop-blur hover:bg-[rgba(185,191,255,0.6)] transition-colors cursor-pointer">
                 <span className="text-xs sm:text-sm font-medium text-black/65">
-                  v0.4.22
+                  v0.4.25
                 </span>
                 <span className="text-xs sm:text-sm font-medium text-black ">
                   —

build_filter_image.sh

@@ -1 +1 @@
-docker build  -f Dockerfile . -t katanemo/plano -t katanemo/plano:0.4.22
+docker build  -f Dockerfile . -t katanemo/plano -t katanemo/plano:0.4.25

cli/planoai/__init__.py

@@ -1,3 +1,3 @@
 """Plano CLI - Intelligent Prompt Gateway."""
 
-__version__ = "0.4.22"
+__version__ = "0.4.25"

cli/planoai/config_generator.py

@@ -39,6 +39,42 @@ CHATGPT_API_BASE = "https://chatgpt.com/backend-api/codex"
 CHATGPT_DEFAULT_ORIGINATOR = "codex_cli_rs"
 CHATGPT_DEFAULT_USER_AGENT = "codex_cli_rs/0.0.0 (Unknown 0; unknown) unknown"
 
+KIMI_CODE_API_HOST = "api.kimi.com"
+KIMI_CODE_DEFAULT_USER_AGENT = "KimiCLI/1.3"
+
+
+def normalize_kimi_code_base_url(base_url: str) -> str:
+    """Ensure Kimi Code API base URLs include the /v1 suffix."""
+    parsed = urlparse(base_url)
+    if parsed.hostname != KIMI_CODE_API_HOST:
+        return base_url
+    path = parsed.path.rstrip("/")
+    if path.endswith("/coding"):
+        return f"{parsed.scheme}://{parsed.netloc}{path}/v1"
+    return base_url
+
+
+def apply_kimi_code_provider_defaults(model_provider: dict) -> None:
+    """Inject Kimi Code API defaults (User-Agent, normalized base URL)."""
+    base_url = model_provider.get("base_url")
+    if not base_url:
+        return
+    parsed = urlparse(base_url)
+    model_id = model_provider.get("model", "")
+    is_kimi_code = (
+        parsed.hostname == KIMI_CODE_API_HOST or model_id == "kimi-for-coding"
+    )
+    if not is_kimi_code:
+        return
+
+    normalized = normalize_kimi_code_base_url(base_url)
+    if normalized != base_url:
+        model_provider["base_url"] = normalized
+
+    headers = model_provider.setdefault("headers", {})
+    headers.setdefault("User-Agent", KIMI_CODE_DEFAULT_USER_AGENT)
+
+
 SUPPORTED_PROVIDERS = (
     SUPPORTED_PROVIDERS_WITHOUT_BASE_URL + SUPPORTED_PROVIDERS_WITH_BASE_URL
 )
@@ -463,6 +499,8 @@ def validate_and_render_schema():
                 headers.setdefault("session_id", str(uuid.uuid4()))
                 model_provider["headers"] = headers
 
+            apply_kimi_code_provider_defaults(model_provider)
+
             updated_model_providers.append(model_provider)
 
             if model_provider.get("base_url", None):
@@ -562,15 +600,15 @@ def validate_and_render_schema():
                     "Please provide model_providers either under listeners or at root level, not both. Currently we don't support multiple listeners with model_providers"
                 )
 
-    # Validate input_filters IDs on listeners reference valid agent/filter IDs
+    # Validate listener-level filter IDs reference valid agent/filter IDs.
     for listener in listeners:
-        listener_input_filters = listener.get("input_filters", [])
-        for fc_id in listener_input_filters:
-            if fc_id not in agent_id_keys:
-                raise Exception(
-                    f"Listener '{listener.get('name', 'unknown')}' references input_filters id '{fc_id}' "
-                    f"which is not defined in agents or filters. Available ids: {', '.join(sorted(agent_id_keys))}"
-                )
+        for filter_field in ("input_filters", "output_filters"):
+            for fc_id in listener.get(filter_field, []):
+                if fc_id not in agent_id_keys:
+                    raise Exception(
+                        f"Listener '{listener.get('name', 'unknown')}' references {filter_field} id '{fc_id}' "
+                        f"which is not defined in agents or filters. Available ids: {', '.join(sorted(agent_id_keys))}"
+                    )
 
     # Validate model aliases if present
     if "model_aliases" in config_yaml:

cli/planoai/consts.py

@@ -5,7 +5,7 @@ PLANO_COLOR = "#969FF4"
 
 SERVICE_NAME_ARCHGW = "plano"
 PLANO_DOCKER_NAME = "plano"
-PLANO_DOCKER_IMAGE = os.getenv("PLANO_DOCKER_IMAGE", "katanemo/plano:0.4.22")
+PLANO_DOCKER_IMAGE = os.getenv("PLANO_DOCKER_IMAGE", "katanemo/plano:0.4.25")
 DEFAULT_OTEL_TRACING_GRPC_ENDPOINT = "http://localhost:4317"
 
 # Native mode constants

cli/planoai/main.py

@@ -7,7 +7,6 @@ import contextlib
 import logging
 import rich_click as click
 import yaml
-from planoai import targets
 from planoai.defaults import (
     DEFAULT_LLM_LISTENER_PORT,
     detect_providers,
@@ -622,28 +621,6 @@ def down(docker, verbose):
                 )
 
 
-@click.command()
-@click.option(
-    "--f",
-    "--file",
-    type=click.Path(exists=True),
-    required=True,
-    help="Path to the Python file",
-)
-def generate_prompt_targets(file):
-    """Generats prompt_targets from python methods.
-    Note: This works for simple data types like ['int', 'float', 'bool', 'str', 'list', 'tuple', 'set', 'dict']:
-    If you have a complex pydantic data type, you will have to flatten those manually until we add support for it.
-    """
-
-    print(f"Processing file: {file}")
-    if not file.endswith(".py"):
-        print("Error: Input file must be a .py file")
-        sys.exit(1)
-
-    targets.generate_prompt_targets(file)
-
-
 @click.command()
 @click.option(
     "--debug",
@@ -741,7 +718,6 @@ main.add_command(down)
 main.add_command(build)
 main.add_command(logs)
 main.add_command(cli_agent)
-main.add_command(generate_prompt_targets)
 main.add_command(init_cmd, name="init")
 main.add_command(trace_cmd, name="trace")
 main.add_command(chatgpt_cmd, name="chatgpt")

cli/planoai/rich_click_config.py

@@ -63,9 +63,5 @@ def configure_rich_click(plano_color: str) -> None:
                 "name": "Observability",
                 "commands": ["trace", "obs"],
             },
-            {
-                "name": "Utilities",
-                "commands": ["generate-prompt-targets"],
-            },
         ],
     }

cli/planoai/targets.py

@@ -1,365 +0,0 @@
-import ast
-import sys
-import yaml
-from typing import Any
-
-FLASK_ROUTE_DECORATORS = ["route", "get", "post", "put", "delete", "patch"]
-FASTAPI_ROUTE_DECORATORS = ["get", "post", "put", "delete", "patch"]
-
-
-def detect_framework(tree: Any) -> str:
-    """Detect whether the file is using Flask or FastAPI based on imports."""
-    for node in ast.walk(tree):
-        if isinstance(node, ast.ImportFrom):
-            if node.module == "flask":
-                return "flask"
-            elif node.module == "fastapi":
-                return "fastapi"
-    return "unknown"
-
-
-def get_route_decorators(node: Any, framework: str) -> list:
-    """Extract route decorators based on the framework."""
-    decorators = []
-    for decorator in node.decorator_list:
-        if isinstance(decorator, ast.Call) and isinstance(
-            decorator.func, ast.Attribute
-        ):
-            if framework == "flask" and decorator.func.attr in FLASK_ROUTE_DECORATORS:
-                decorators.append(decorator.func.attr)
-            elif (
-                framework == "fastapi"
-                and decorator.func.attr in FASTAPI_ROUTE_DECORATORS
-            ):
-                decorators.append(decorator.func.attr)
-    return decorators
-
-
-def get_route_path(node: Any, framework: str) -> str:
-    """Extract route path based on the framework."""
-    for decorator in node.decorator_list:
-        if isinstance(decorator, ast.Call) and decorator.args:
-            return decorator.args[0].s  # Assuming it's a string literal
-
-
-def is_pydantic_model(annotation: ast.expr, tree: ast.AST) -> bool:
-    """Check if a given type annotation is a Pydantic model."""
-    # We walk through the AST to find class definitions and check if they inherit from Pydantic's BaseModel
-    if isinstance(annotation, ast.Name):
-        for node in ast.walk(tree):
-            if isinstance(node, ast.ClassDef) and node.name == annotation.id:
-                for base in node.bases:
-                    if isinstance(base, ast.Name) and base.id == "BaseModel":
-                        return True
-    return False
-
-
-def get_pydantic_model_fields(model_name: str, tree: ast.AST) -> list:
-    """Extract fields from a Pydantic model, handling list, tuple, set, dict types, and direct default values."""
-    fields = []
-
-    for node in ast.walk(tree):
-        if isinstance(node, ast.ClassDef) and node.name == model_name:
-            for stmt in node.body:
-                if isinstance(stmt, ast.AnnAssign):
-                    # Initialize the default field description
-                    field_type = "Unknown: Please Fix This!"
-                    description = "Field, description not present. Please fix."
-                    default_value = None
-                    required = True  # Assume the field is required initially
-
-                    # Check if the field uses Field() with required status and description
-                    if (
-                        stmt.value
-                        and isinstance(stmt.value, ast.Call)
-                        and isinstance(stmt.value.func, ast.Name)
-                        and stmt.value.func.id == "Field"
-                    ):
-                        # Extract the description argument inside the Field call
-                        for keyword in stmt.value.keywords:
-                            if keyword.arg == "description" and isinstance(
-                                keyword.value, ast.Str
-                            ):
-                                description = keyword.value.s
-                            if keyword.arg == "default":
-                                default_value = keyword.value
-                        # If Ellipsis (...) is used, it means the field is required
-                        if (
-                            stmt.value.args
-                            and isinstance(stmt.value.args[0], ast.Constant)
-                            and stmt.value.args[0].value is Ellipsis
-                        ):
-                            required = True
-                        else:
-                            required = False
-
-                    # Handle direct default values (e.g., name: str = "John Doe")
-                    elif stmt.value is not None:
-                        if isinstance(stmt.value, ast.Constant):
-                            # Set the default value from the assignment (e.g., name: str = "John Doe")
-                            default_value = stmt.value.value
-                            required = (
-                                False  # Not required since it has a default value
-                            )
-
-                    # Always extract the field type, even if there's a default value
-                    if isinstance(stmt.annotation, ast.Subscript):
-                        # Get the base type (list, tuple, set, dict)
-                        base_type = (
-                            stmt.annotation.value.id
-                            if isinstance(stmt.annotation.value, ast.Name)
-                            else "Unknown"
-                        )
-
-                        # Handle only list, tuple, set, dict and ignore the inner types
-                        if base_type.lower() in ["list", "tuple", "set", "dict"]:
-                            field_type = base_type.lower()
-
-                    # Handle the ellipsis '...' for required fields if no Field() call
-                    elif (
-                        isinstance(stmt.value, ast.Constant)
-                        and stmt.value.value is Ellipsis
-                    ):
-                        required = True
-
-                    # Handle simple types like str, int, etc.
-                    if isinstance(stmt.annotation, ast.Name):
-                        field_type = stmt.annotation.id
-
-                    field_info = {
-                        "name": stmt.target.id,
-                        "type": field_type,  # Always set the field type
-                        "description": description,
-                        "default": default_value,  # Handle direct default values
-                        "required": required,
-                    }
-                    fields.append(field_info)
-
-    return fields
-
-
-def get_function_parameters(node: ast.FunctionDef, tree: ast.AST) -> list:
-    """Extract the parameters and their types from the function definition."""
-    parameters = []
-
-    # Extract docstring to find descriptions
-    docstring = ast.get_docstring(node)
-    arg_descriptions = extract_arg_descriptions_from_docstring(docstring)
-
-    # Extract default values
-    defaults = [None] * (
-        len(node.args.args) - len(node.args.defaults)
-    ) + node.args.defaults  # Align defaults with args
-    for arg, default in zip(node.args.args, defaults):
-        if arg.arg != "self":  # Skip 'self' or 'cls' in class methods
-            param_info = {
-                "name": arg.arg,
-                "description": arg_descriptions.get(arg.arg, "[ADD DESCRIPTION]"),
-            }
-
-            # Handle Pydantic model types
-            if hasattr(arg, "annotation") and is_pydantic_model(arg.annotation, tree):
-                # Extract and flatten Pydantic model fields
-                pydantic_fields = get_pydantic_model_fields(arg.annotation.id, tree)
-                parameters.extend(
-                    pydantic_fields
-                )  # Flatten the model fields into the parameters list
-                continue  # Skip adding the current param_info for the model since we expand the fields
-
-            # Handle standard Python types (int, float, str, etc.)
-            elif hasattr(arg, "annotation") and isinstance(arg.annotation, ast.Name):
-                if arg.annotation.id in [
-                    "int",
-                    "float",
-                    "bool",
-                    "str",
-                    "list",
-                    "tuple",
-                    "set",
-                    "dict",
-                ]:
-                    param_info["type"] = arg.annotation.id
-                else:
-                    param_info["type"] = "[UNKNOWN - PLEASE FIX]"
-
-            # Handle generic subscript types (e.g., Optional, List[Type], etc.)
-            elif hasattr(arg, "annotation") and isinstance(
-                arg.annotation, ast.Subscript
-            ):
-                if isinstance(
-                    arg.annotation.value, ast.Name
-                ) and arg.annotation.value.id in ["list", "tuple", "set", "dict"]:
-                    param_info["type"] = (
-                        f"{arg.annotation.value.id}"  # e.g., "List", "Tuple", etc.
-                    )
-                else:
-                    param_info["type"] = "[UNKNOWN - PLEASE FIX]"
-
-            # Default for unknown types
-            else:
-                param_info["type"] = (
-                    "[UNKNOWN - PLEASE FIX]"  # If unable to detect type
-                )
-
-            # Handle default values
-            if default is not None:
-                if isinstance(default, ast.Constant) or isinstance(
-                    default, ast.NameConstant
-                ):
-                    param_info["default"] = (
-                        default.value
-                    )  # Use the default value directly
-                else:
-                    param_info["default"] = "[UNKNOWN DEFAULT]"  # Unknown default type
-                param_info["required"] = False  # Optional since it has a default value
-            else:
-                param_info["default"] = None
-                param_info["required"] = True  # Required if no default value
-
-            parameters.append(param_info)
-
-    return parameters
-
-
-def get_function_docstring(node: Any) -> str:
-    """Extract the function's docstring description if present."""
-    # Check if the first node is a docstring
-    if isinstance(node.body[0], ast.Expr) and isinstance(node.body[0].value, ast.Str):
-        # Get the entire docstring
-        full_docstring = node.body[0].value.s.strip()
-
-        # Split the docstring by double newlines (to separate description from fields like Args)
-        description = full_docstring.split("\n\n")[0].strip()
-
-        return description
-
-    return "No description provided."
-
-
-def extract_arg_descriptions_from_docstring(docstring: str) -> dict:
-    """Extract descriptions for function parameters from the 'Args' section of the docstring."""
-    descriptions = {}
-    if not docstring:
-        return descriptions
-
-    in_args_section = False
-    current_param = None
-    for line in docstring.splitlines():
-        line = line.strip()
-
-        # Detect the start of the 'Args' section
-        if line.startswith("Args:"):
-            in_args_section = True
-            continue  # Proceed to the next line after 'Args:'
-
-        # End of 'Args' section if no indentation and no colon
-        if in_args_section and not line.startswith(" ") and ":" not in line:
-            break  # Stop processing if we reach a new section
-
-        # Process lines in the 'Args' section
-        if in_args_section:
-            if ":" in line:
-                # Extract parameter name and description
-                param_name, description = line.split(":", 1)
-                descriptions[param_name.strip()] = description.strip()
-                current_param = param_name.strip()
-            elif current_param and line.startswith(" "):
-                # Handle multiline descriptions (indented lines)
-                descriptions[current_param] += f" {line.strip()}"
-
-    return descriptions
-
-
-def generate_prompt_targets(input_file_path: str) -> None:
-    """Introspect routes and generate YAML for either Flask or FastAPI."""
-    with open(input_file_path, "r") as source:
-        tree = ast.parse(source.read())
-
-    # Detect the framework (Flask or FastAPI)
-    framework = detect_framework(tree)
-    if framework == "unknown":
-        print("Could not detect Flask or FastAPI in the file.")
-        return
-
-    # Extract routes
-    routes = []
-    for node in ast.walk(tree):
-        if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):
-            route_decorators = get_route_decorators(node, framework)
-            if route_decorators:
-                route_path = get_route_path(node, framework)
-                function_params = get_function_parameters(
-                    node, tree
-                )  # Get parameters for the route
-                function_docstring = get_function_docstring(node)  # Extract docstring
-                routes.append(
-                    {
-                        "name": node.name,
-                        "path": route_path,
-                        "methods": route_decorators,
-                        "parameters": function_params,  # Add parameters to the route
-                        "description": function_docstring,  # Add the docstring as the description
-                    }
-                )
-
-    # Generate YAML structure
-    output_structure = {"prompt_targets": []}
-
-    for route in routes:
-        target = {
-            "name": route["name"],
-            "endpoint": [
-                {
-                    "name": "app_server",
-                    "path": route["path"],
-                }
-            ],
-            "description": route["description"],  # Use extracted docstring
-            "parameters": [
-                {
-                    "name": param["name"],
-                    "type": param["type"],
-                    "description": f"{param['description']}",
-                    **(
-                        {"default": param["default"]}
-                        if "default" in param and param["default"] is not None
-                        else {}
-                    ),  # Only add default if it's set
-                    "required": param["required"],
-                }
-                for param in route["parameters"]
-            ],
-        }
-
-        if route["name"] == "default":
-            # Special case for `information_extraction` based on your YAML format
-            target["type"] = "default"
-            target["auto-llm-dispatch-on-response"] = True
-
-        output_structure["prompt_targets"].append(target)
-
-    # Output as YAML
-    print(
-        yaml.dump(output_structure, sort_keys=False, default_flow_style=False, indent=3)
-    )
-
-
-if __name__ == "__main__":
-    if len(sys.argv) != 2:
-        print("Usage: python targets.py <input_file>")
-        sys.exit(1)
-
-    input_file = sys.argv[1]
-
-    # Automatically generate the output file name
-    if input_file.endswith(".py"):
-        output_file = input_file.replace(".py", "_prompt_targets.yml")
-    else:
-        print("Error: Input file must be a .py file")
-        sys.exit(1)
-
-    # Call the function with the input and generated output file names
-    generate_prompt_targets(input_file, output_file)
-
-# Example usage:
-# python targets.py api.yaml

cli/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "planoai"
-version = "0.4.22"
+version = "0.4.25"
 description = "Python-based CLI tool to manage Plano."
 authors = [{name = "Katanemo Labs, Inc."}]
 readme = "README.md"

cli/test/test_config_generator.py

@@ -3,8 +3,10 @@ import pytest
 import yaml
 from unittest import mock
 from planoai.config_generator import (
-    validate_and_render_schema,
+    apply_kimi_code_provider_defaults,
     migrate_inline_routing_preferences,
+    normalize_kimi_code_base_url,
+    validate_and_render_schema,
 )
 
 
@@ -327,6 +329,63 @@ routing_preferences:
 tracing:
   random_sampling: 100
 
+""",
+    },
+    {
+        "id": "unknown_listener_output_filter",
+        "expected_error": "references output_filters id 'missing_output_guard'",
+        "plano_config": """
+version: v0.4.0
+
+filters:
+  - id: input_guard
+    url: http://localhost:10500
+    type: http
+
+listeners:
+  - name: llm
+    type: model
+    port: 12000
+    input_filters:
+      - input_guard
+    output_filters:
+      - missing_output_guard
+
+model_providers:
+  - model: openai/gpt-4o-mini
+    access_key: $OPENAI_API_KEY
+    default: true
+
+""",
+    },
+    {
+        "id": "valid_listener_output_filter",
+        "expected_error": None,
+        "plano_config": """
+version: v0.4.0
+
+filters:
+  - id: input_guard
+    url: http://localhost:10500
+    type: http
+  - id: output_guard
+    url: http://localhost:10501
+    type: http
+
+listeners:
+  - name: llm
+    type: model
+    port: 12000
+    input_filters:
+      - input_guard
+    output_filters:
+      - output_guard
+
+model_providers:
+  - model: openai/gpt-4o-mini
+    access_key: $OPENAI_API_KEY
+    default: true
+
 """,
     },
 ]
@@ -738,3 +797,29 @@ model_providers:
     migrate_inline_routing_preferences(config_yaml)
 
     assert config_yaml["version"] == "v0.5.0"
+
+
+def test_normalize_kimi_code_base_url_appends_v1_suffix():
+    assert (
+        normalize_kimi_code_base_url("https://api.kimi.com/coding")
+        == "https://api.kimi.com/coding/v1"
+    )
+    assert (
+        normalize_kimi_code_base_url("https://api.kimi.com/coding/")
+        == "https://api.kimi.com/coding/v1"
+    )
+    assert (
+        normalize_kimi_code_base_url("https://api.kimi.com/coding/v1")
+        == "https://api.kimi.com/coding/v1"
+    )
+
+
+def test_apply_kimi_code_provider_defaults_injects_user_agent():
+    provider = {
+        "model": "kimi-for-coding",
+        "base_url": "https://api.kimi.com/coding",
+        "access_key": "$MOONSHOTAI_API_KEY",
+    }
+    apply_kimi_code_provider_defaults(provider)
+    assert provider["base_url"] == "https://api.kimi.com/coding/v1"
+    assert provider["headers"]["User-Agent"] == "KimiCLI/1.3"

cli/uv.lock

@@ -337,7 +337,7 @@ wheels = [
 
 [[package]]
 name = "planoai"
-version = "0.4.22"
+version = "0.4.25"
 source = { editable = "." }
 dependencies = [
     { name = "click" },

config/plano_config_schema.yaml

@@ -194,6 +194,7 @@ properties:
             - digitalocean
             - vercel
             - openrouter
+            - moonshotai
         headers:
           type: object
           additionalProperties:
@@ -252,6 +253,7 @@ properties:
             - digitalocean
             - vercel
             - openrouter
+            - moonshotai
         headers:
           type: object
           additionalProperties:

crates/Cargo.lock

@@ -2552,9 +2552,9 @@ dependencies = [
 
 [[package]]
 name = "proxy-wasm"
-version = "0.2.4"
+version = "0.2.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f8d35d9e2bc5104e2e954b149aa1d5f9fa3bb27f73b45b2706020fed101db685"
+checksum = "de8f6564bd52c2f4ff79fa5d1bd3bc10d8f822162af8d527e121e46703496aa0"
 dependencies = [
  "hashbrown 0.16.1",
  "log",
@@ -2752,12 +2752,18 @@ dependencies = [
  "num-bigint",
  "percent-encoding",
  "pin-project-lite",
+ "rustls 0.23.38",
+ "rustls-native-certs 0.7.3",
+ "rustls-pemfile 2.2.0",
+ "rustls-pki-types",
  "ryu",
  "sha1_smol",
  "socket2 0.5.10",
  "tokio",
+ "tokio-rustls 0.26.4",
  "tokio-util",
  "url",
+ "webpki-roots 0.26.11",
 ]
 
 [[package]]
@@ -2965,7 +2971,20 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "a9aace74cb666635c918e9c12bc0d348266037aa8eb599b5cba565709a8dff00"
 dependencies = [
  "openssl-probe 0.1.6",
- "rustls-pemfile",
+ "rustls-pemfile 1.0.4",
+ "schannel",
+ "security-framework 2.11.1",
+]
+
+[[package]]
+name = "rustls-native-certs"
+version = "0.7.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e5bfb394eeed242e909609f56089eecfe5fda225042e8b171791b9c95f5931e5"
+dependencies = [
+ "openssl-probe 0.1.6",
+ "rustls-pemfile 2.2.0",
+ "rustls-pki-types",
  "schannel",
  "security-framework 2.11.1",
 ]
@@ -2991,6 +3010,15 @@ dependencies = [
  "base64 0.21.7",
 ]
 
+[[package]]
+name = "rustls-pemfile"
+version = "2.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "dce314e5fee3f39953d46bb63bb8a46d40c2f8fb7cc5a3b6cab2bde9721d6e50"
+dependencies = [
+ "rustls-pki-types",
+]
+
 [[package]]
 name = "rustls-pki-types"
 version = "1.14.0"
@@ -4024,7 +4052,7 @@ dependencies = [
  "serde_json",
  "ureq-proto",
  "utf8-zero",
- "webpki-roots",
+ "webpki-roots 1.0.6",
 ]
 
 [[package]]
@@ -4278,6 +4306,15 @@ dependencies = [
  "wasm-bindgen",
 ]
 
+[[package]]
+name = "webpki-roots"
+version = "0.26.11"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "521bc38abb08001b01866da9f51eb7c5d647a19260e00054a8c7fd5f9e57f7a9"
+dependencies = [
+ "webpki-roots 1.0.6",
+]
+
 [[package]]
 name = "webpki-roots"
 version = "1.0.6"

crates/brightstaff/Cargo.toml

@@ -43,7 +43,7 @@ lru = "0.12"
 metrics = "0.23"
 metrics-exporter-prometheus = { version = "0.15", default-features = false, features = ["http-listener"] }
 metrics-process = "2.1"
-redis = { version = "0.27", features = ["tokio-comp"] }
+redis = { version = "0.27", features = ["tokio-comp", "tokio-rustls-comp", "tls-rustls-webpki-roots"] }
 reqwest = { version = "0.12.15", features = ["stream"] }
 serde = { version = "1.0.219", features = ["derive"] }
 serde_json = "1.0.140"

crates/brightstaff/src/main.rs

@@ -142,25 +142,19 @@ async fn init_app_state(
         .listeners
         .iter()
         .find(|l| l.listener_type == ListenerType::Model);
-    let resolve_chain = |filter_ids: Option<Vec<String>>| -> Option<ResolvedFilterChain> {
-        filter_ids.map(|ids| {
-            let agents = ids
-                .iter()
-                .filter_map(|id| {
-                    global_agent_map
-                        .get(id)
-                        .map(|a: &Agent| (id.clone(), a.clone()))
-                })
-                .collect();
-            ResolvedFilterChain {
-                filter_ids: ids,
-                agents,
-            }
-        })
-    };
     let filter_pipeline = Arc::new(FilterPipeline {
-        input: resolve_chain(model_listener.and_then(|l| l.input_filters.clone())),
-        output: resolve_chain(model_listener.and_then(|l| l.output_filters.clone())),
+        input: resolve_filter_chain(
+            "input_filters",
+            model_listener.and_then(|l| l.input_filters.clone()),
+            &global_agent_map,
+        )
+        .map_err(|e| format!("failed to resolve model listener input filters: {e}"))?,
+        output: resolve_filter_chain(
+            "output_filters",
+            model_listener.and_then(|l| l.output_filters.clone()),
+            &global_agent_map,
+        )
+        .map_err(|e| format!("failed to resolve model listener output filters: {e}"))?,
     });
 
     let overrides = config.overrides.clone().unwrap_or_default();
@@ -350,6 +344,29 @@ async fn init_app_state(
     })
 }
 
+fn resolve_filter_chain(
+    field_name: &str,
+    filter_ids: Option<Vec<String>>,
+    global_agent_map: &HashMap<String, Agent>,
+) -> Result<Option<ResolvedFilterChain>, String> {
+    let Some(ids) = filter_ids else {
+        return Ok(None);
+    };
+
+    let mut agents = HashMap::new();
+    for id in &ids {
+        let agent = global_agent_map
+            .get(id)
+            .ok_or_else(|| format!("{field_name} id '{id}' is not defined in agents or filters"))?;
+        agents.insert(id.clone(), agent.clone());
+    }
+
+    Ok(Some(ResolvedFilterChain {
+        filter_ids: ids,
+        agents,
+    }))
+}
+
 /// Initialize the conversation state storage backend (if configured).
 async fn init_state_storage(
     config: &Configuration,
@@ -588,3 +605,63 @@ async fn main() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
     let state = Arc::new(init_app_state(&config).await?);
     run_server(state).await
 }
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    fn test_agent(id: &str) -> Agent {
+        Agent {
+            id: id.to_string(),
+            transport: None,
+            tool: None,
+            url: "http://localhost:10500".to_string(),
+            agent_type: Some("http".to_string()),
+        }
+    }
+
+    #[test]
+    fn resolve_filter_chain_keeps_valid_filter_references() {
+        let agent = test_agent("output_guard");
+        let global_agent_map = HashMap::from([(agent.id.clone(), agent)]);
+
+        let resolved = resolve_filter_chain(
+            "output_filters",
+            Some(vec!["output_guard".to_string()]),
+            &global_agent_map,
+        )
+        .expect("filter chain should resolve")
+        .expect("filter chain should be present");
+
+        assert_eq!(resolved.filter_ids, vec!["output_guard".to_string()]);
+        assert!(resolved.agents.contains_key("output_guard"));
+    }
+
+    #[test]
+    fn resolve_filter_chain_errors_on_missing_output_filter_reference() {
+        let global_agent_map = HashMap::new();
+
+        let err = resolve_filter_chain(
+            "output_filters",
+            Some(vec!["missing_output_guard".to_string()]),
+            &global_agent_map,
+        )
+        .expect_err("missing output filter should fail closed");
+
+        assert!(err.contains("output_filters id 'missing_output_guard'"));
+    }
+
+    #[test]
+    fn resolve_filter_chain_errors_on_missing_input_filter_reference() {
+        let global_agent_map = HashMap::new();
+
+        let err = resolve_filter_chain(
+            "input_filters",
+            Some(vec!["missing_input_guard".to_string()]),
+            &global_agent_map,
+        )
+        .expect_err("missing input filter should fail closed");
+
+        assert!(err.contains("input_filters id 'missing_input_guard'"));
+    }
+}

crates/common/src/api/open_ai.rs

@@ -53,7 +53,7 @@ impl Serialize for FunctionParameters {
     where
         S: serde::Serializer,
     {
-        // select all requried parameters
+        // select all required parameters
         let required: Vec<&String> = self
             .properties
             .iter()

crates/common/src/configuration.rs

@@ -400,6 +400,10 @@ pub enum LlmProviderType {
     Vercel,
     #[serde(rename = "openrouter")]
     OpenRouter,
+    #[serde(rename = "astraflow")]
+    Astraflow,
+    #[serde(rename = "astraflow_cn")]
+    AstraflowCN,
 }
 
 impl Display for LlmProviderType {
@@ -425,6 +429,8 @@ impl Display for LlmProviderType {
             LlmProviderType::DigitalOcean => write!(f, "digitalocean"),
             LlmProviderType::Vercel => write!(f, "vercel"),
             LlmProviderType::OpenRouter => write!(f, "openrouter"),
+            LlmProviderType::Astraflow => write!(f, "astraflow"),
+            LlmProviderType::AstraflowCN => write!(f, "astraflow_cn"),
         }
     }
 }

crates/hermesllm/src/apis/anthropic.rs

@@ -128,6 +128,7 @@ pub struct MessagesRequest {
 pub enum MessagesRole {
     User,
     Assistant,
+    System,
 }
 
 /// Cache control types for content blocks
@@ -632,6 +633,7 @@ impl MessagesRole {
         match self {
             MessagesRole::User => "user",
             MessagesRole::Assistant => "assistant",
+            MessagesRole::System => "system",
         }
     }
 }

crates/hermesllm/src/apis/openai.rs

@@ -1,3 +1,4 @@
+use log::warn;
 use serde::{Deserialize, Serialize};
 use serde_json::Value;
 use serde_with::skip_serializing_none;
@@ -136,6 +137,37 @@ impl ChatCompletionsRequest {
             self.temperature = Some(1.0);
         }
     }
+
+    /// Strip request fields that Kimi Code API (`kimi-for-coding`) rejects or mishandles.
+    pub fn normalize_for_kimi_code_api(&mut self) {
+        if self.stream_options.is_some() {
+            warn!("kimi-for-coding: stripping unsupported stream_options from upstream request");
+            self.stream_options = None;
+        }
+        if self.reasoning_effort.is_some() {
+            warn!("kimi-for-coding: stripping unsupported reasoning_effort from upstream request");
+            self.reasoning_effort = None;
+        }
+        if self.web_search_options.is_some() {
+            warn!(
+                "kimi-for-coding: stripping unsupported web_search_options from upstream request"
+            );
+            self.web_search_options = None;
+        }
+        if self.service_tier.is_some() {
+            warn!("kimi-for-coding: stripping unsupported service_tier from upstream request");
+            self.service_tier = None;
+        }
+        if self.store.is_some() {
+            warn!("kimi-for-coding: stripping unsupported store from upstream request");
+            self.store = None;
+        }
+    }
+}
+
+/// True when the upstream model id is Moonshot's Kimi Code endpoint model.
+pub fn is_kimi_code_model(model: &str) -> bool {
+    model == "kimi-for-coding"
 }
 
 // ============================================================================

crates/hermesllm/src/bin/fetch_models.rs

@@ -1,12 +1,21 @@
-// Fetch latest provider models from canonical provider APIs and update provider_models.yaml
+// Fetch latest provider models from canonical provider APIs and merge into
+// provider_models.yaml.
+//
+// Behavior is non-destructive: only providers we successfully fetch this run
+// are replaced. Providers whose API key is missing, or whose fetch fails, are
+// left untouched in the existing file. This means partial runs (e.g. without
+// AWS or Google creds) can't accidentally wipe out provider entries you don't
+// have keys for locally.
+//
 // Usage:
-//   Optional: OPENAI_API_KEY, ANTHROPIC_API_KEY, DEEPSEEK_API_KEY, GROK_API_KEY,
-//             DASHSCOPE_API_KEY, MOONSHOT_API_KEY, ZHIPU_API_KEY, GOOGLE_API_KEY
-//   Required: AWS CLI configured for Amazon Bedrock models
-//   cargo run --bin fetch_models
+//   Optional: OPENAI_API_KEY, ANTHROPIC_API_KEY, MISTRAL_API_KEY,
+//             DEEPSEEK_API_KEY, GROK_API_KEY, DASHSCOPE_API_KEY,
+//             MOONSHOT_API_KEY, ZHIPU_API_KEY, MIMO_API_KEY, GOOGLE_API_KEY
+//   Optional: AWS CLI configured for Amazon Bedrock models
+//   cargo run --bin fetch_models --features model-fetch
 
 use serde::{Deserialize, Serialize};
-use std::collections::HashMap;
+use std::collections::BTreeMap;
 
 fn main() {
     // Default to writing in the same directory as this source file
@@ -19,16 +28,33 @@ fn main() {
         .nth(1)
         .unwrap_or_else(|| default_path.to_string_lossy().to_string());
 
-    println!("Fetching latest models from provider APIs...");
+    println!("Loading existing {}...", output_path);
+    let existing = match load_existing_models(&output_path) {
+        Ok(map) => {
+            if map.is_empty() {
+                println!("  (none — starting fresh)");
+            } else {
+                println!("  loaded {} existing providers", map.len());
+            }
+            map
+        }
+        Err(e) => {
+            eprintln!("Error loading existing {}: {}", output_path, e);
+            eprintln!("Refusing to overwrite a file we can't parse. Fix or delete it and re-run.");
+            std::process::exit(1);
+        }
+    };
 
-    match fetch_all_models() {
+    println!("\nFetching latest models from provider APIs...");
+
+    match fetch_all_models(existing) {
         Ok(models) => {
             let yaml = serde_yaml::to_string(&models).expect("Failed to serialize models");
 
             std::fs::write(&output_path, yaml).expect("Failed to write provider_models.yaml");
 
             println!(
-                "✓ Successfully updated {} providers ({} models) to {}",
+                "✓ Wrote {} providers ({} models) to {}",
                 models.metadata.total_providers, models.metadata.total_models, output_path
             );
         }
@@ -44,6 +70,18 @@ fn main() {
     }
 }
 
+fn load_existing_models(
+    path: &str,
+) -> Result<BTreeMap<String, Vec<String>>, Box<dyn std::error::Error>> {
+    let content = match std::fs::read_to_string(path) {
+        Ok(c) => c,
+        Err(e) if e.kind() == std::io::ErrorKind::NotFound => return Ok(BTreeMap::new()),
+        Err(e) => return Err(Box::new(e)),
+    };
+    let parsed: ProviderModels = serde_yaml::from_str(&content)?;
+    Ok(parsed.providers)
+}
+
 // OpenAI-compatible API response (used by most providers)
 #[derive(Debug, Deserialize)]
 struct OpenAICompatibleModel {
@@ -68,21 +106,36 @@ struct GoogleResponse {
     models: Vec<GoogleModel>,
 }
 
-#[derive(Debug, Serialize)]
+#[derive(Debug, Serialize, Deserialize)]
 struct ProviderModels {
+    #[serde(default = "default_version")]
     version: String,
+    #[serde(default = "default_source")]
     source: String,
-    providers: HashMap<String, Vec<String>>,
+    #[serde(default)]
+    providers: BTreeMap<String, Vec<String>>,
+    #[serde(default)]
     metadata: Metadata,
 }
 
-#[derive(Debug, Serialize)]
+#[derive(Debug, Default, Serialize, Deserialize)]
 struct Metadata {
+    #[serde(default)]
     total_providers: usize,
+    #[serde(default)]
     total_models: usize,
+    #[serde(default)]
     last_updated: String,
 }
 
+fn default_version() -> String {
+    "1.0".to_string()
+}
+
+fn default_source() -> String {
+    "canonical-apis".to_string()
+}
+
 fn is_text_model(model_id: &str) -> bool {
     let id_lower = model_id.to_lowercase();
 
@@ -273,8 +326,13 @@ fn fetch_bedrock_amazon_models() -> Result<Vec<String>, Box<dyn std::error::Erro
     Ok(amazon_models)
 }
 
-fn fetch_all_models() -> Result<ProviderModels, Box<dyn std::error::Error>> {
-    let mut providers: HashMap<String, Vec<String>> = HashMap::new();
+fn fetch_all_models(
+    existing: BTreeMap<String, Vec<String>>,
+) -> Result<ProviderModels, Box<dyn std::error::Error>> {
+    let mut providers = existing;
+    let mut updated: Vec<String> = Vec::new();
+    let mut skipped: Vec<String> = Vec::new();
+    let mut failed: Vec<String> = Vec::new();
     let mut errors: Vec<String> = Vec::new();
 
     // Configuration: provider name, env var, API URL, prefix for model IDs
@@ -324,90 +382,131 @@ fn fetch_all_models() -> Result<ProviderModels, Box<dyn std::error::Error>> {
         ),
     ];
 
+    // Helper that records the outcome of a fetch attempt and only mutates
+    // `providers` on success, so missing/failed providers keep their existing
+    // entries (or stay absent if there were none).
+    let mut record =
+        |name: &str,
+         env_var: Option<&str>,
+         result: Option<Result<Vec<String>, Box<dyn std::error::Error>>>,
+         providers: &mut BTreeMap<String, Vec<String>>| match result {
+            Some(Ok(models)) => {
+                println!("  ✓ {}: {} models", name, models.len());
+                providers.insert(name.to_string(), models);
+                updated.push(name.to_string());
+            }
+            Some(Err(e)) => {
+                let kept = providers
+                    .get(name)
+                    .map(|v| format!(" (keeping existing {} models)", v.len()))
+                    .unwrap_or_default();
+                let err_msg = format!("  ✗ {}: {}{}", name, e, kept);
+                eprintln!("{}", err_msg);
+                errors.push(err_msg);
+                failed.push(name.to_string());
+            }
+            None => {
+                let kept = providers
+                    .get(name)
+                    .map(|v| format!(" (keeping existing {} models)", v.len()))
+                    .unwrap_or_else(|| " (no existing entry)".to_string());
+                let label = env_var
+                    .map(|v| format!("{} not set", v))
+                    .unwrap_or_else(|| "no credentials".to_string());
+                println!("  ⊘ {}: {}{}", name, label, kept);
+                skipped.push(name.to_string());
+            }
+        };
+
     // Fetch from OpenAI-compatible providers
     for (provider_name, env_var, api_url, prefix) in provider_configs {
-        if let Ok(api_key) = std::env::var(env_var) {
-            match fetch_openai_compatible_models(api_url, &api_key, prefix) {
-                Ok(models) => {
-                    println!("  ✓ {}: {} models", provider_name, models.len());
-                    providers.insert(provider_name.to_string(), models);
-                }
-                Err(e) => {
-                    let err_msg = format!("  ✗ {}: {}", provider_name, e);
-                    eprintln!("{}", err_msg);
-                    errors.push(err_msg);
-                }
-            }
-        } else {
-            println!("  ⊘ {}: {} not set (skipped)", provider_name, env_var);
-        }
+        let result = std::env::var(env_var)
+            .ok()
+            .map(|api_key| fetch_openai_compatible_models(api_url, &api_key, prefix));
+        record(provider_name, Some(env_var), result, &mut providers);
     }
 
     // Fetch Anthropic models (different authentication)
-    if let Ok(api_key) = std::env::var("ANTHROPIC_API_KEY") {
-        match fetch_anthropic_models(&api_key) {
-            Ok(models) => {
-                println!("  ✓ anthropic: {} models", models.len());
-                providers.insert("anthropic".to_string(), models);
-            }
-            Err(e) => {
-                let err_msg = format!("  ✗ anthropic: {}", e);
-                eprintln!("{}", err_msg);
-                errors.push(err_msg);
-            }
-        }
-    } else {
-        println!("  ⊘ anthropic: ANTHROPIC_API_KEY not set (skipped)");
-    }
+    let anthropic_result = std::env::var("ANTHROPIC_API_KEY")
+        .ok()
+        .map(|key| fetch_anthropic_models(&key));
+    record(
+        "anthropic",
+        Some("ANTHROPIC_API_KEY"),
+        anthropic_result,
+        &mut providers,
+    );
 
     // Fetch Google models (different API format)
-    if let Ok(api_key) = std::env::var("GOOGLE_API_KEY") {
-        match fetch_google_models(&api_key) {
-            Ok(models) => {
-                println!("  ✓ google: {} models", models.len());
-                providers.insert("google".to_string(), models);
-            }
-            Err(e) => {
-                let err_msg = format!("  ✗ google: {}", e);
-                eprintln!("{}", err_msg);
-                errors.push(err_msg);
-            }
-        }
-    } else {
-        println!("  ⊘ google: GOOGLE_API_KEY not set (skipped)");
-    }
+    let google_result = std::env::var("GOOGLE_API_KEY")
+        .ok()
+        .map(|key| fetch_google_models(&key));
+    record(
+        "google",
+        Some("GOOGLE_API_KEY"),
+        google_result,
+        &mut providers,
+    );
 
-    // Fetch Amazon models from AWS Bedrock
-    match fetch_bedrock_amazon_models() {
-        Ok(models) => {
-            println!("  ✓ amazon: {} models (via AWS Bedrock)", models.len());
-            providers.insert("amazon".to_string(), models);
-        }
-        Err(e) => {
-            let err_msg = format!("  ✗ amazon: {} (AWS Bedrock required)", e);
-            eprintln!("{}", err_msg);
-            errors.push(err_msg);
-        }
-    }
+    // Fetch Amazon models from AWS Bedrock. Only attempt if the AWS CLI is on
+    // PATH and any AWS credential is configured — otherwise treat as skipped
+    // so we don't drop the existing amazon entry on machines / CI runs without
+    // Bedrock access.
+    let amazon_result = if aws_credentials_available() {
+        Some(fetch_bedrock_amazon_models())
+    } else {
+        None
+    };
+    record(
+        "amazon",
+        Some("AWS credentials"),
+        amazon_result,
+        &mut providers,
+    );
 
     if providers.is_empty() {
-        return Err("No models fetched from any provider. Check API keys.".into());
+        return Err(
+            "No existing data and no models fetched. Set at least one API key and re-run.".into(),
+        );
     }
 
     let total_providers = providers.len();
     let total_models: usize = providers.values().map(|v| v.len()).sum();
 
+    println!("\nSummary:");
     println!(
-        "\n✅ Successfully fetched models from {} providers",
-        total_providers
+        "  updated: {} ({})",
+        updated.len(),
+        if updated.is_empty() {
+            "none".to_string()
+        } else {
+            updated.join(", ")
+        }
     );
-    if !errors.is_empty() {
-        println!("⚠️  {} providers failed", errors.len());
+    println!(
+        "  skipped (kept existing): {} ({})",
+        skipped.len(),
+        if skipped.is_empty() {
+            "none".to_string()
+        } else {
+            skipped.join(", ")
+        }
+    );
+    if !failed.is_empty() {
+        println!(
+            "  failed (kept existing): {} ({})",
+            failed.len(),
+            failed.join(", ")
+        );
     }
+    println!(
+        "✅ Final state: {} providers, {} models",
+        total_providers, total_models
+    );
 
     Ok(ProviderModels {
-        version: "1.0".to_string(),
-        source: "canonical-apis".to_string(),
+        version: default_version(),
+        source: default_source(),
         providers,
         metadata: Metadata {
             total_providers,
@@ -416,3 +515,10 @@ fn fetch_all_models() -> Result<ProviderModels, Box<dyn std::error::Error>> {
         },
     })
 }
+
+fn aws_credentials_available() -> bool {
+    std::env::var("AWS_ACCESS_KEY_ID").is_ok()
+        || std::env::var("AWS_PROFILE").is_ok()
+        || std::env::var("AWS_SESSION_TOKEN").is_ok()
+        || std::env::var("AWS_WEB_IDENTITY_TOKEN_FILE").is_ok()
+}

crates/hermesllm/src/bin/provider_models.yaml

@@ -13,6 +13,77 @@ providers:
   - amazon/amazon.nova-premier-v1:0
   - amazon/amazon.nova-lite-v1:0
   - amazon/amazon.nova-micro-v1:0
+  anthropic:
+  - anthropic/claude-fable-5
+  - anthropic/claude-opus-4-8
+  - anthropic/claude-opus-4-7
+  - anthropic/claude-sonnet-4-6
+  - anthropic/claude-opus-4-6
+  - anthropic/claude-opus-4-5-20251101
+  - anthropic/claude-opus-4-5
+  - anthropic/claude-haiku-4-5-20251001
+  - anthropic/claude-haiku-4-5
+  - anthropic/claude-sonnet-4-5-20250929
+  - anthropic/claude-sonnet-4-5
+  - anthropic/claude-opus-4-1-20250805
+  - anthropic/claude-opus-4-1
+  - anthropic/claude-opus-4-20250514
+  - anthropic/claude-opus-4
+  - anthropic/claude-sonnet-4-20250514
+  - anthropic/claude-sonnet-4
+  chatgpt:
+  - chatgpt/gpt-5.4
+  - chatgpt/gpt-5.3-codex
+  - chatgpt/gpt-5.2
+  deepseek:
+  - deepseek/deepseek-v4-flash
+  - deepseek/deepseek-v4-pro
+  digitalocean:
+  - digitalocean/openai-gpt-4.1
+  - digitalocean/openai-gpt-4o
+  - digitalocean/openai-gpt-4o-mini
+  - digitalocean/openai-gpt-5
+  - digitalocean/openai-gpt-5-mini
+  - digitalocean/openai-gpt-5-nano
+  - digitalocean/openai-gpt-5.1-codex-max
+  - digitalocean/openai-gpt-5.2
+  - digitalocean/openai-gpt-5.2-pro
+  - digitalocean/openai-gpt-5.3-codex
+  - digitalocean/openai-gpt-5.4
+  - digitalocean/openai-gpt-5.4-mini
+  - digitalocean/openai-gpt-5.4-nano
+  - digitalocean/openai-gpt-5.4-pro
+  - digitalocean/openai-gpt-oss-120b
+  - digitalocean/openai-gpt-oss-20b
+  - digitalocean/openai-o1
+  - digitalocean/openai-o3
+  - digitalocean/openai-o3-mini
+  - digitalocean/anthropic-claude-4.1-opus
+  - digitalocean/anthropic-claude-4.5-sonnet
+  - digitalocean/anthropic-claude-4.6-sonnet
+  - digitalocean/anthropic-claude-haiku-4.5
+  - digitalocean/anthropic-claude-opus-4
+  - digitalocean/anthropic-claude-opus-4.5
+  - digitalocean/anthropic-claude-opus-4.6
+  - digitalocean/anthropic-claude-opus-4.7
+  - digitalocean/anthropic-claude-sonnet-4
+  - digitalocean/alibaba-qwen3-32b
+  - digitalocean/arcee-trinity-large-thinking
+  - digitalocean/deepseek-3.2
+  - digitalocean/deepseek-r1-distill-llama-70b
+  - digitalocean/gemma-4-31B-it
+  - digitalocean/glm-5
+  - digitalocean/kimi-k2.5
+  - digitalocean/llama3.3-70b-instruct
+  - digitalocean/minimax-m2.5
+  - digitalocean/nvidia-nemotron-3-super-120b
+  - digitalocean/qwen3-coder-flash
+  - digitalocean/qwen3.5-397b-a17b
+  - digitalocean/all-mini-lm-l6-v2
+  - digitalocean/gte-large-en-v1.5
+  - digitalocean/multi-qa-mpnet-base-dot-v1
+  - digitalocean/qwen3-embedding-0.6b
+  - digitalocean/router:software-engineering
   google:
   - google/gemini-2.5-flash
   - google/gemini-2.5-pro
@@ -22,12 +93,6 @@ providers:
   - google/gemini-2.0-flash-lite
   - google/gemini-2.5-flash-preview-tts
   - google/gemini-2.5-pro-preview-tts
-  - google/gemma-3-1b-it
-  - google/gemma-3-4b-it
-  - google/gemma-3-12b-it
-  - google/gemma-3-27b-it
-  - google/gemma-3n-e4b-it
-  - google/gemma-3n-e2b-it
   - google/gemma-4-26b-a4b-it
   - google/gemma-4-31b-it
   - google/gemini-flash-latest
@@ -40,13 +105,22 @@ providers:
   - google/gemini-3.1-pro-preview
   - google/gemini-3.1-pro-preview-customtools
   - google/gemini-3.1-flash-lite-preview
+  - google/gemini-3.1-flash-lite
   - google/gemini-3-pro-image-preview
+  - google/gemini-3-pro-image
   - google/nano-banana-pro-preview
   - google/gemini-3.1-flash-image-preview
+  - google/gemini-3.1-flash-image
+  - google/gemini-3.5-flash
   - google/lyria-3-clip-preview
   - google/lyria-3-pro-preview
+  - google/gemini-3.1-flash-tts-preview
   - google/gemini-robotics-er-1.5-preview
+  - google/gemini-robotics-er-1.6-preview
   - google/gemini-2.5-computer-use-preview-10-2025
+  - google/antigravity-preview-05-2026
+  - google/deep-research-max-preview-04-2026
+  - google/deep-research-preview-04-2026
   - google/deep-research-pro-preview-12-2025
   mistralai:
   - mistralai/mistral-medium-2505
@@ -60,183 +134,62 @@ providers:
   - mistralai/mistral-tiny-latest
   - mistralai/codestral-2508
   - mistralai/codestral-latest
+  - mistralai/mistral-code-latest
+  - mistralai/mistral-code-fim-latest
   - mistralai/devstral-2512
-  - mistralai/mistral-vibe-cli-latest
   - mistralai/devstral-medium-latest
   - mistralai/devstral-latest
+  - mistralai/mistral-code-agent-latest
   - mistralai/mistral-small-2603
   - mistralai/mistral-small-latest
   - mistralai/mistral-vibe-cli-fast
-  - mistralai/mistral-small-2506
+  - mistralai/magistral-small-latest
   - mistralai/magistral-medium-2509
   - mistralai/magistral-medium-latest
-  - mistralai/magistral-small-2509
-  - mistralai/magistral-small-latest
   - mistralai/labs-leanstral-2603
   - mistralai/mistral-large-2512
   - mistralai/mistral-large-latest
+  - mistralai/mistral-large-2512
+  - mistralai/mistral-large-latest
   - mistralai/ministral-3b-2512
   - mistralai/ministral-3b-latest
   - mistralai/ministral-8b-2512
   - mistralai/ministral-8b-latest
   - mistralai/ministral-14b-2512
   - mistralai/ministral-14b-latest
-  - mistralai/mistral-large-2411
-  - mistralai/pixtral-large-2411
-  - mistralai/pixtral-large-latest
-  - mistralai/mistral-large-pixtral-2411
-  - mistralai/devstral-small-2507
-  - mistralai/devstral-medium-2507
-  - mistralai/labs-mistral-small-creative
+  - mistralai/mistral-medium-3-5
+  - mistralai/mistral-medium-3.5
+  - mistralai/mistral-medium-3
+  - mistralai/mistral-medium-2604
+  - mistralai/mistral-medium-c21211-r0-75
+  - mistralai/mistral-vibe-cli-latest
+  - mistralai/mistral-medium-3-5
+  - mistralai/mistral-medium-3.5
+  - mistralai/mistral-medium-3
+  - mistralai/mistral-medium-2604
+  - mistralai/mistral-medium-c21211-r0-75
+  - mistralai/mistral-vibe-cli-latest
+  - mistralai/magistral-small-2509
+  - mistralai/mistral-small-2506
   - mistralai/mistral-embed-2312
   - mistralai/mistral-embed
   - mistralai/codestral-embed
   - mistralai/codestral-embed-2505
-  anthropic:
-  - anthropic/claude-sonnet-4-6
-  - anthropic/claude-opus-4-6
-  - anthropic/claude-opus-4-7
-  - anthropic/claude-opus-4-5-20251101
-  - anthropic/claude-opus-4-5
-  - anthropic/claude-haiku-4-5-20251001
-  - anthropic/claude-haiku-4-5
-  - anthropic/claude-sonnet-4-5-20250929
-  - anthropic/claude-sonnet-4-5
-  - anthropic/claude-opus-4-1-20250805
-  - anthropic/claude-opus-4-1
-  - anthropic/claude-opus-4-20250514
-  - anthropic/claude-opus-4
-  - anthropic/claude-sonnet-4-20250514
-  - anthropic/claude-sonnet-4
-  - anthropic/claude-3-haiku-20240307
-  - anthropic/claude-3-haiku
-  qwen:
-  - qwen/qwen3.6-plus-2026-04-02
-  - qwen/qwen3.6-plus
-  - qwen/wan2.7-image
-  - qwen/deepseek-v3.2
-  - qwen/qwen3-asr-flash-2026-02-10
-  - qwen/qwen3.5-flash-2026-02-23
-  - qwen/qwen3.5-flash
-  - qwen/qwen3.5-122b-a10b
-  - qwen/qwen3.5-35b-a3b
-  - qwen/qwen3.5-27b
-  - qwen/qwen3-coder-next
-  - qwen/qwen3.5-397b-a17b
-  - qwen/qwen3.5-plus-2026-02-15
-  - qwen/qwen3.5-plus
-  - qwen/qwen3-vl-flash-2026-01-22
-  - qwen/qwen3-max-2026-01-23
-  - qwen/qwen-plus-character
-  - qwen/qwen-flash-character
-  - qwen/qwen-flash
-  - qwen/qwen3-vl-plus-2025-12-19
-  - qwen/qwen3-omni-flash-2025-12-01
-  - qwen/qwen3-livetranslate-flash-2025-12-01
-  - qwen/qwen3-livetranslate-flash
-  - qwen/qwen-mt-lite
-  - qwen/qwen-plus-2025-12-01
-  - qwen/qwen-mt-flash
-  - qwen/ccai-pro
-  - qwen/tongyi-tingwu-slp
-  - qwen/qwen3-vl-flash
-  - qwen/qwen3-vl-flash-2025-10-15
-  - qwen/qwen3-omni-flash
-  - qwen/qwen3-omni-flash-2025-09-15
-  - qwen/qwen3-omni-30b-a3b-captioner
-  - qwen/qwen2.5-7b-instruct
-  - qwen/qwen2.5-14b-instruct
-  - qwen/qwen2.5-32b-instruct
-  - qwen/qwen2.5-72b-instruct
-  - qwen/qwen2.5-14b-instruct-1m
-  - qwen/qwen2.5-7b-instruct-1m
-  - qwen/qwen-max-2025-01-25
-  - qwen/qwen-max-latest
-  - qwen/qwen-turbo-2024-11-01
-  - qwen/qwen-turbo-latest
-  - qwen/qwen-plus-latest
-  - qwen/qwen-plus-2025-01-25
-  - qwen/qwq-plus-2025-03-05
-  - qwen/qwen-mt-turbo
-  - qwen/qwen-mt-plus
-  - qwen/qwen-coder-plus
-  - qwen/qwq-plus
-  - qwen/qwen2.5-vl-32b-instruct
-  - qwen/qvq-max
-  - qwen/qwen-omni-turbo
-  - qwen/qwen3-8b
-  - qwen/qwen3-30b-a3b
-  - qwen/qwen3-235b-a22b
-  - qwen/qwen-turbo-2025-04-28
-  - qwen/qwen-plus-2025-04-28
-  - qwen/qwen-vl-max-2025-04-08
-  - qwen/qwen-vl-plus-2025-01-25
-  - qwen/qwen-vl-plus-latest
-  - qwen/qwen-vl-max-latest
-  - qwen/qwen-vl-plus-2025-05-07
-  - qwen/qwen3-coder-plus
-  - qwen/qwen3-coder-480b-a35b-instruct
-  - qwen/qwen3-235b-a22b-instruct-2507
-  - qwen/qwen-plus-2025-07-14
-  - qwen/qwen3-coder-plus-2025-07-22
-  - qwen/qwen3-235b-a22b-thinking-2507
-  - qwen/qwen3-coder-flash
-  - qwen/qwen-vl-max
-  - qwen/qwen-vl-max-2025-08-13
-  - qwen/qwen3-max
-  - qwen/qwen3-max-2025-09-23
-  - qwen/qwen3-vl-plus
-  - qwen/qwen3-vl-235b-a22b-instruct
-  - qwen/qwen3-vl-235b-a22b-thinking
-  - qwen/qwen3-30b-a3b-thinking-2507
-  - qwen/qwen3-30b-a3b-instruct-2507
-  - qwen/qwen3-14b
-  - qwen/qwen3-32b
-  - qwen/qwen3-0.6b
-  - qwen/qwen3-4b
-  - qwen/qwen3-1.7b
-  - qwen/qwen-vl-plus
-  - qwen/qwen3-coder-plus-2025-09-23
-  - qwen/qwen3-vl-plus-2025-09-23
-  - qwen/qwen-plus-2025-09-11
-  - qwen/qwen3-next-80b-a3b-thinking
-  - qwen/qwen3-next-80b-a3b-instruct
-  - qwen/qwen3-max-preview
-  - qwen/qwen2-7b-instruct
-  - qwen/qwen-max
-  - qwen/qwen-plus
-  - qwen/qwen-turbo
-  z-ai:
-  - z-ai/glm-4.5
-  - z-ai/glm-4.5-air
-  - z-ai/glm-4.6
-  - z-ai/glm-4.7
-  - z-ai/glm-5
-  - z-ai/glm-5-turbo
-  - z-ai/glm-5.1
-  x-ai:
-  - x-ai/grok-3
-  - x-ai/grok-3-mini
-  - x-ai/grok-4-0709
-  - x-ai/grok-4-1-fast-non-reasoning
-  - x-ai/grok-4-1-fast-reasoning
-  - x-ai/grok-4-fast-non-reasoning
-  - x-ai/grok-4-fast-reasoning
-  - x-ai/grok-4.20-0309-non-reasoning
-  - x-ai/grok-4.20-0309-reasoning
-  - x-ai/grok-4.20-multi-agent-0309
-  - x-ai/grok-code-fast-1
-  - x-ai/grok-imagine-image
-  - x-ai/grok-imagine-video
+  moonshotai:
+  - moonshotai/kimi-k2.5
+  - moonshotai/kimi-k2.6
+  - moonshotai/moonshot-v1-32k
+  - moonshotai/moonshot-v1-8k
+  - moonshotai/moonshot-v1-128k-vision-preview
+  - moonshotai/moonshot-v1-auto
+  - moonshotai/moonshot-v1-8k-vision-preview
+  - moonshotai/moonshot-v1-128k
+  - moonshotai/moonshot-v1-32k-vision-preview
   openai:
+  - openai/gpt-3.5-turbo
+  - openai/gpt-3.5-turbo-16k
   - openai/gpt-4-0613
   - openai/gpt-4
-  - openai/gpt-3.5-turbo
-  - openai/gpt-5.4-mini
-  - openai/gpt-5.4
-  - openai/gpt-5.4-nano-2026-03-17
-  - openai/gpt-5.4-nano
-  - openai/gpt-5.4-mini-2026-03-17
   - openai/gpt-3.5-turbo-instruct
   - openai/gpt-3.5-turbo-instruct-0914
   - openai/gpt-3.5-turbo-1106
@@ -306,80 +259,137 @@ providers:
   - openai/gpt-5.4-2026-03-05
   - openai/gpt-5.4-pro
   - openai/gpt-5.4-pro-2026-03-05
-  - openai/gpt-3.5-turbo-16k
+  - openai/gpt-5.4
+  - openai/gpt-5.4-nano-2026-03-17
+  - openai/gpt-5.4-nano
+  - openai/gpt-5.4-mini-2026-03-17
+  - openai/gpt-5.4-mini
+  - openai/gpt-5.5
+  - openai/gpt-5.5-2026-04-23
+  - openai/gpt-5.5-pro
+  - openai/gpt-5.5-pro-2026-04-23
+  - openai/chat-latest
   - openai/ft:gpt-3.5-turbo-0613:katanemo::8CMZbm0P
-  deepseek:
-  - deepseek/deepseek-chat
-  - deepseek/deepseek-reasoner
-  moonshotai:
-  - moonshotai/kimi-k2-thinking
-  - moonshotai/moonshot-v1-auto
-  - moonshotai/moonshot-v1-32k-vision-preview
-  - moonshotai/moonshot-v1-128k
-  - moonshotai/kimi-k2-turbo-preview
-  - moonshotai/kimi-k2-0905-preview
-  - moonshotai/moonshot-v1-128k-vision-preview
-  - moonshotai/moonshot-v1-32k
-  - moonshotai/moonshot-v1-8k-vision-preview
-  - moonshotai/kimi-k2.5
-  - moonshotai/moonshot-v1-8k
-  - moonshotai/kimi-k2-thinking-turbo
-  - moonshotai/kimi-k2-0711-preview
+  qwen:
+  - qwen/qwen3.7-plus-2026-05-26
+  - qwen/qwen3.7-plus
+  - qwen/kimi-k2.6
+  - qwen/glm-5.1
+  - qwen/qwen3.7-max-2026-05-17
+  - qwen/qwen3.7-max-preview
+  - qwen/qwen3.7-max-2026-05-20
+  - qwen/qwen3.7-max
+  - qwen/deepseek-v4-flash
+  - qwen/deepseek-v4-pro
+  - qwen/qwen3.6-27b
+  - qwen/qwen3.5-plus-2026-04-20
+  - qwen/qwen3.6-max-preview
+  - qwen/qwen3.6-35b-a3b
+  - qwen/qwen3.6-flash
+  - qwen/qwen3.6-flash-2026-04-16
+  - qwen/qwen3.5-omni-plus-2026-03-15
+  - qwen/qwen3.5-omni-plus
+  - qwen/qwen3.5-omni-flash-2026-03-15
+  - qwen/qwen3.5-omni-flash
+  - qwen/qwen3.6-plus-2026-04-02
+  - qwen/qwen3.6-plus
+  - qwen/wan2.7-image
+  - qwen/deepseek-v3.2
+  - qwen/qwen3-asr-flash-2026-02-10
+  - qwen/qwen3.5-flash-2026-02-23
+  - qwen/qwen3.5-flash
+  - qwen/qwen3.5-122b-a10b
+  - qwen/qwen3.5-35b-a3b
+  - qwen/qwen3.5-27b
+  - qwen/qwen3-coder-next
+  - qwen/qwen3.5-397b-a17b
+  - qwen/qwen3.5-plus-2026-02-15
+  - qwen/qwen3.5-plus
+  - qwen/qwen3-vl-flash-2026-01-22
+  - qwen/qwen3-max-2026-01-23
+  - qwen/qwen-plus-character
+  - qwen/qwen-flash-character
+  - qwen/qwen-flash
+  - qwen/qwen3-vl-plus-2025-12-19
+  - qwen/qwen3-omni-flash-2025-12-01
+  - qwen/qwen3-livetranslate-flash-2025-12-01
+  - qwen/qwen3-livetranslate-flash
+  - qwen/qwen-mt-lite
+  - qwen/qwen-plus-2025-12-01
+  - qwen/qwen-mt-flash
+  - qwen/ccai-pro
+  - qwen/tongyi-tingwu-slp
+  - qwen/qwen3-vl-flash
+  - qwen/qwen3-vl-flash-2025-10-15
+  - qwen/qwen3-omni-flash
+  - qwen/qwen3-omni-flash-2025-09-15
+  - qwen/qwen3-omni-30b-a3b-captioner
+  - qwen/qwen-plus-latest
+  - qwen/qwen-plus-2025-01-25
+  - qwen/qwq-plus-2025-03-05
+  - qwen/qwen-mt-turbo
+  - qwen/qwen-mt-plus
+  - qwen/qwen-coder-plus
+  - qwen/qwq-plus
+  - qwen/qvq-max
+  - qwen/qwen-omni-turbo
+  - qwen/qwen3-8b
+  - qwen/qwen3-30b-a3b
+  - qwen/qwen3-235b-a22b
+  - qwen/qwen-plus-2025-04-28
+  - qwen/qwen3-coder-plus
+  - qwen/qwen3-coder-480b-a35b-instruct
+  - qwen/qwen3-235b-a22b-instruct-2507
+  - qwen/qwen-plus-2025-07-14
+  - qwen/qwen3-coder-plus-2025-07-22
+  - qwen/qwen3-235b-a22b-thinking-2507
+  - qwen/qwen3-coder-flash
+  - qwen/qwen-vl-max
+  - qwen/qwen3-max
+  - qwen/qwen3-max-2025-09-23
+  - qwen/qwen3-vl-plus
+  - qwen/qwen3-vl-235b-a22b-instruct
+  - qwen/qwen3-vl-235b-a22b-thinking
+  - qwen/qwen3-30b-a3b-thinking-2507
+  - qwen/qwen3-30b-a3b-instruct-2507
+  - qwen/qwen3-14b
+  - qwen/qwen3-32b
+  - qwen/qwen-vl-plus
+  - qwen/qwen3-coder-plus-2025-09-23
+  - qwen/qwen3-vl-plus-2025-09-23
+  - qwen/qwen-plus-2025-09-11
+  - qwen/qwen3-next-80b-a3b-thinking
+  - qwen/qwen3-next-80b-a3b-instruct
+  - qwen/qwen3-max-preview
+  - qwen/qwen2-7b-instruct
+  - qwen/qwen-max
+  - qwen/qwen-plus
+  - qwen/qwen-turbo
+  x-ai:
+  - x-ai/grok-4.20-0309-non-reasoning
+  - x-ai/grok-4.20-0309-reasoning
+  - x-ai/grok-4.20-multi-agent-0309
+  - x-ai/grok-4.3
+  - x-ai/grok-build-0.1
+  - x-ai/grok-imagine-image
+  - x-ai/grok-imagine-video
+  - x-ai/grok-imagine-video-1.5-preview
   xiaomi:
   - xiaomi/mimo-v2-flash
   - xiaomi/mimo-v2-omni
   - xiaomi/mimo-v2-pro
-  chatgpt:
-  - chatgpt/gpt-5.4
-  - chatgpt/gpt-5.3-codex
-  - chatgpt/gpt-5.2
-  digitalocean:
-  - digitalocean/openai-gpt-4.1
-  - digitalocean/openai-gpt-4o
-  - digitalocean/openai-gpt-4o-mini
-  - digitalocean/openai-gpt-5
-  - digitalocean/openai-gpt-5-mini
-  - digitalocean/openai-gpt-5-nano
-  - digitalocean/openai-gpt-5.1-codex-max
-  - digitalocean/openai-gpt-5.2
-  - digitalocean/openai-gpt-5.2-pro
-  - digitalocean/openai-gpt-5.3-codex
-  - digitalocean/openai-gpt-5.4
-  - digitalocean/openai-gpt-5.4-mini
-  - digitalocean/openai-gpt-5.4-nano
-  - digitalocean/openai-gpt-5.4-pro
-  - digitalocean/openai-gpt-oss-120b
-  - digitalocean/openai-gpt-oss-20b
-  - digitalocean/openai-o1
-  - digitalocean/openai-o3
-  - digitalocean/openai-o3-mini
-  - digitalocean/anthropic-claude-4.1-opus
-  - digitalocean/anthropic-claude-4.5-sonnet
-  - digitalocean/anthropic-claude-4.6-sonnet
-  - digitalocean/anthropic-claude-haiku-4.5
-  - digitalocean/anthropic-claude-opus-4
-  - digitalocean/anthropic-claude-opus-4.5
-  - digitalocean/anthropic-claude-opus-4.6
-  - digitalocean/anthropic-claude-opus-4.7
-  - digitalocean/anthropic-claude-sonnet-4
-  - digitalocean/alibaba-qwen3-32b
-  - digitalocean/arcee-trinity-large-thinking
-  - digitalocean/deepseek-3.2
-  - digitalocean/deepseek-r1-distill-llama-70b
-  - digitalocean/gemma-4-31B-it
-  - digitalocean/glm-5
-  - digitalocean/kimi-k2.5
-  - digitalocean/llama3.3-70b-instruct
-  - digitalocean/minimax-m2.5
-  - digitalocean/nvidia-nemotron-3-super-120b
-  - digitalocean/qwen3-coder-flash
-  - digitalocean/qwen3.5-397b-a17b
-  - digitalocean/all-mini-lm-l6-v2
-  - digitalocean/gte-large-en-v1.5
-  - digitalocean/multi-qa-mpnet-base-dot-v1
-  - digitalocean/qwen3-embedding-0.6b
-  - digitalocean/router:software-engineering
+  - xiaomi/mimo-v2.5
+  - xiaomi/mimo-v2.5-asr
+  - xiaomi/mimo-v2.5-pro
+  z-ai:
+  - z-ai/glm-4.5
+  - z-ai/glm-4.5-air
+  - z-ai/glm-4.6
+  - z-ai/glm-4.7
+  - z-ai/glm-5
+  - z-ai/glm-5-turbo
+  - z-ai/glm-5.1
 metadata:
   total_providers: 13
-  total_models: 364
-  last_updated: 2026-04-20T00:00:00.000000+00:00
+  total_models: 375
+  last_updated: 2026-06-09T22:50:12.186709+00:00

crates/hermesllm/src/clients/endpoints.rs

@@ -500,6 +500,19 @@ mod tests {
             "/custom/api/v2/chat/completions"
         );
 
+        // Kimi Code API: base_url path prefix already includes /coding/v1
+        assert_eq!(
+            api.target_endpoint_for_provider(
+                &ProviderId::Moonshotai,
+                "/v1/messages",
+                "kimi-for-coding",
+                false,
+                Some("/coding/v1"),
+                false
+            ),
+            "/coding/v1/chat/completions"
+        );
+
         // Test Groq with custom prefix
         assert_eq!(
             api.target_endpoint_for_provider(

crates/hermesllm/src/providers/id.rs

@@ -48,6 +48,8 @@ pub enum ProviderId {
     DigitalOcean,
     Vercel,
     OpenRouter,
+    Astraflow,
+    AstraflowCN,
 }
 
 impl TryFrom<&str> for ProviderId {
@@ -81,6 +83,8 @@ impl TryFrom<&str> for ProviderId {
             "do_ai" => Ok(ProviderId::DigitalOcean), // alias
             "vercel" => Ok(ProviderId::Vercel),
             "openrouter" => Ok(ProviderId::OpenRouter),
+            "astraflow" => Ok(ProviderId::Astraflow),
+            "astraflow_cn" => Ok(ProviderId::AstraflowCN),
             _ => Err(format!("Unknown provider: {}", value)),
         }
     }
@@ -107,6 +111,7 @@ impl ProviderId {
             ProviderId::Qwen => "qwen",
             ProviderId::ChatGPT => "chatgpt",
             ProviderId::DigitalOcean => "digitalocean",
+            ProviderId::Astraflow | ProviderId::AstraflowCN => return Vec::new(),
             _ => return Vec::new(),
         };
 
@@ -174,7 +179,9 @@ impl ProviderId {
                 | ProviderId::Qwen
                 | ProviderId::DigitalOcean
                 | ProviderId::OpenRouter
-                | ProviderId::ChatGPT,
+                | ProviderId::ChatGPT
+                | ProviderId::Astraflow
+                | ProviderId::AstraflowCN,
                 SupportedAPIsFromClient::AnthropicMessagesAPI(_),
             ) => SupportedUpstreamAPIs::OpenAIChatCompletions(OpenAIApi::ChatCompletions),
 
@@ -196,7 +203,9 @@ impl ProviderId {
                 | ProviderId::Qwen
                 | ProviderId::DigitalOcean
                 | ProviderId::OpenRouter
-                | ProviderId::ChatGPT,
+                | ProviderId::ChatGPT
+                | ProviderId::Astraflow
+                | ProviderId::AstraflowCN,
                 SupportedAPIsFromClient::OpenAIChatCompletions(_),
             ) => SupportedUpstreamAPIs::OpenAIChatCompletions(OpenAIApi::ChatCompletions),
 
@@ -267,6 +276,8 @@ impl Display for ProviderId {
             ProviderId::DigitalOcean => write!(f, "digitalocean"),
             ProviderId::Vercel => write!(f, "vercel"),
             ProviderId::OpenRouter => write!(f, "openrouter"),
+            ProviderId::Astraflow => write!(f, "astraflow"),
+            ProviderId::AstraflowCN => write!(f, "astraflow_cn"),
         }
     }
 }

crates/hermesllm/src/providers/request.rs

@@ -1,5 +1,6 @@
 use crate::apis::anthropic::MessagesRequest;
-use crate::apis::openai::ChatCompletionsRequest;
+use crate::apis::openai::{is_kimi_code_model, ChatCompletionsRequest};
+use log::warn;
 
 use crate::apis::amazon_bedrock::{ConverseRequest, ConverseStreamRequest};
 use crate::apis::openai_responses::ResponsesAPIRequest;
@@ -90,6 +91,24 @@ impl ProviderRequestType {
             }
         }
 
+        if matches!(
+            upstream_api,
+            SupportedUpstreamAPIs::OpenAIChatCompletions(_)
+        ) {
+            if let Self::ChatCompletionsRequest(req) = self {
+                if is_kimi_code_model(req.model()) {
+                    req.normalize_for_kimi_code_api();
+                }
+            } else if let Self::MessagesRequest(req) = self {
+                if is_kimi_code_model(req.model.as_str()) && req.thinking.is_some() {
+                    warn!(
+                        "kimi-for-coding: stripping unsupported thinking config from upstream request"
+                    );
+                    req.thinking = None;
+                }
+            }
+        }
+
         // ChatGPT requires instructions, store=false, and input as a list
         if provider_id == ProviderId::ChatGPT {
             if let Self::ResponsesAPIRequest(req) = self {
@@ -879,6 +898,42 @@ mod tests {
         assert!(req.web_search_options.is_none());
     }
 
+    #[test]
+    fn test_normalize_for_upstream_kimi_code_strips_unsupported_chat_fields() {
+        use crate::apis::openai::{Message, MessageContent, OpenAIApi, Role, StreamOptions};
+
+        let mut request = ProviderRequestType::ChatCompletionsRequest(ChatCompletionsRequest {
+            model: "kimi-for-coding".to_string(),
+            messages: vec![Message {
+                role: Role::User,
+                content: Some(MessageContent::Text("hello".to_string())),
+                name: None,
+                tool_calls: None,
+                tool_call_id: None,
+            }],
+            stream_options: Some(StreamOptions {
+                include_usage: Some(true),
+            }),
+            reasoning_effort: Some("high".to_string()),
+            web_search_options: Some(serde_json::json!({"search_context_size":"medium"})),
+            ..Default::default()
+        });
+
+        request
+            .normalize_for_upstream(
+                ProviderId::Moonshotai,
+                &SupportedUpstreamAPIs::OpenAIChatCompletions(OpenAIApi::ChatCompletions),
+            )
+            .unwrap();
+
+        let ProviderRequestType::ChatCompletionsRequest(req) = request else {
+            panic!("expected chat request");
+        };
+        assert!(req.stream_options.is_none());
+        assert!(req.reasoning_effort.is_none());
+        assert!(req.web_search_options.is_none());
+    }
+
     #[test]
     fn test_normalize_for_upstream_non_xai_keeps_chat_web_search_options() {
         use crate::apis::openai::{Message, MessageContent, OpenAIApi, Role};

crates/hermesllm/src/transforms/request/from_anthropic.rs

@@ -223,6 +223,7 @@ impl From<MessagesRole> for Role {
         match val {
             MessagesRole::User => Role::User,
             MessagesRole::Assistant => Role::Assistant,
+            MessagesRole::System => Role::System,
         }
     }
 }
@@ -340,6 +341,11 @@ impl TryFrom<MessagesMessage> for BedrockMessage {
         let role = match message.role {
             MessagesRole::User => ConversationRole::User,
             MessagesRole::Assistant => ConversationRole::Assistant,
+            MessagesRole::System => {
+                return Err(TransformError::UnsupportedConversion(
+                    "System messages must be set via the system prompt, not messages".to_string(),
+                ));
+            }
         };
 
         let mut content_blocks = Vec::new();

demos/llm_routing/preference_based_routing/README.md

@@ -3,7 +3,7 @@ This demo shows how you can use user preferences to route user prompts to approp
 
 ## How to start the demo
 
-Make sure you have Plano CLI installed (`pip install planoai==0.4.22` or `uv tool install planoai==0.4.22`).
+Make sure you have Plano CLI installed (`pip install planoai==0.4.25` or `uv tool install planoai==0.4.25`).
 
 ```bash
 cd demos/llm_routing/preference_based_routing

docs/source/concepts/llm_providers/supported_providers.rst

@@ -432,6 +432,9 @@ Moonshot AI
    * - Model Name
      - Model ID for Config
      - Description
+   * - Kimi for Coding
+     - ``moonshotai/kimi-for-coding``
+     - Kimi Code API model for agentic coding (use with ``base_url: https://api.kimi.com/coding/v1``)
    * - Kimi K2 Preview
      - ``moonshotai/kimi-k2-0905-preview``
      - Foundation model optimized for agentic tasks with 32B activated parameters
@@ -447,6 +450,13 @@ Moonshot AI
 .. code-block:: yaml
 
     llm_providers:
+      # Kimi Code API (Claude Code / agentic clients via Plano translation)
+      - model: moonshotai/kimi-for-coding
+        access_key: $MOONSHOTAI_API_KEY
+        base_url: https://api.kimi.com/coding/v1
+        headers:
+          User-Agent: "KimiCLI/1.3"
+
       # Latest K2 models for agentic tasks
       - model: moonshotai/kimi-k2-0905-preview
         access_key: $MOONSHOTAI_API_KEY

docs/source/concepts/prompt_target.rst

@@ -2,6 +2,15 @@
 
 Prompt Target
 =============
+
+.. deprecated:: v0.4.22
+   **Prompt Targets are deprecated and no longer actively maintained.** This concept is
+   retained for existing users on older Plano configurations, but new applications should
+   not adopt it. For deterministic, task-specific workloads, use :ref:`Agents <agents>`
+   together with :ref:`Function Calling <function_calling>` instead. The
+   ``prompt_targets`` configuration block and related CLI commands will continue to
+   function for now, but may be removed in a future release.
+
 A Prompt Target is a deterministic, task-specific backend function or API endpoint that your application calls via Plano.
 Unlike agents (which handle wide-ranging, open-ended tasks), prompt targets are designed for focused, specific workloads where Plano can add value through input clarification and validation.
 

docs/source/conf.py

@@ -17,7 +17,7 @@ from sphinxawesome_theme.postprocess import Icons
 project = "Plano Docs"
 copyright = "2026, Katanemo Labs, a DigitalOcean Company"
 author = "Katanemo Labs, Inc"
-release = " v0.4.22"
+release = " v0.4.25"
 
 # -- General configuration ---------------------------------------------------
 # https://www.sphinx-doc.org/en/master/usage/configuration.html#general-configuration

docs/source/get_started/overview.rst

@@ -57,10 +57,10 @@ Deep dive into essential ideas and mechanisms behind Plano:
 
         Explore Plano's LLM integration options
 
-    .. grid-item-card:: :octicon:`workflow` Prompt Target
+    .. grid-item-card:: :octicon:`workflow` Prompt Target (Deprecated)
         :link: ../concepts/prompt_target.html
 
-        Understand how Plano handles prompts
+        Deprecated — kept for existing users. New apps should use Agents.
 
 
 Guides

docs/source/get_started/quickstart.rst

@@ -43,7 +43,7 @@ Plano's CLI allows you to manage and interact with the Plano efficiently. To ins
 
 .. code-block:: console
 
-   $ uv tool install planoai==0.4.22
+   $ uv tool install planoai==0.4.25
 
 **Option 2: Install with pip (Traditional)**
 
@@ -51,7 +51,7 @@ Plano's CLI allows you to manage and interact with the Plano efficiently. To ins
 
    $ python -m venv venv
    $ source venv/bin/activate   # On Windows, use: venv\Scripts\activate
-   $ pip install planoai==0.4.22
+   $ pip install planoai==0.4.25
 
 
 .. _llm_routing_quickstart:
@@ -247,6 +247,11 @@ You can then ask a follow-up like "Also book me a hotel near JFK" and Plano-Orch
 Deterministic API calls with prompt targets
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
+.. deprecated:: v0.4.22
+   :ref:`Prompt Targets <prompt_target>` are deprecated and no longer actively
+   maintained. The walkthrough below is preserved for users on existing configs;
+   new applications should use :ref:`Agents <agents>` instead.
+
 Next, we'll show Plano's deterministic API calling using a single prompt target. We'll build a currency exchange backend powered by `https://api.frankfurter.dev/`, assuming USD as the base currency.
 
 Step 1. Create plano config file

docs/source/guides/function_calling.rst

@@ -6,6 +6,12 @@ Function Calling
 **Function Calling** is a powerful feature in Plano that allows your application to dynamically execute backend functions or services based on user prompts.
 This enables seamless integration between natural language interactions and backend operations, turning user inputs into actionable results.
 
+.. deprecated:: v0.4.22
+   The prompt-target based workflow shown below (see :ref:`Step 2 <function_calling>`)
+   is deprecated. :ref:`Prompt Targets <prompt_target>` are no longer actively
+   maintained and may be removed in a future release. For new function-calling
+   workloads, prefer :ref:`Agents <agents>` with tool definitions.
+
 
 What is Function Calling?
 -------------------------

docs/source/resources/cli_reference.rst

@@ -16,7 +16,6 @@ Quick Navigation
 - :ref:`cli_reference_logs`
 - :ref:`cli_reference_init`
 - :ref:`cli_reference_trace`
-- :ref:`cli_reference_prompt_targets`
 - :ref:`cli_reference_cli_agent`
 
 
@@ -260,24 +259,6 @@ Inspect request traces from the local OTLP listener.
 - ``--list`` cannot be combined with a specific trace-id target.
 
 
-.. _cli_reference_prompt_targets:
-
-planoai prompt_targets
-----------------------
-
-Generate prompt-target metadata from Python methods.
-
-**Synopsis**
-
-.. code-block:: console
-
-   $ planoai prompt_targets --file <python-file>
-
-**Options**
-
-- ``--file, --f <python-file>``: required path to a ``.py`` source file.
-
-
 .. _cli_reference_cli_agent:
 
 planoai cli_agent

docs/source/resources/configuration_reference.rst

@@ -7,6 +7,29 @@ The following is a complete reference of the ``plano_config.yml`` that controls
 the Plano gateway. This where you enable capabilities like routing to upstream LLm providers, defining prompt_targets
 where prompts get routed to, apply guardrails, and enable critical agent observability features.
 
+Model provider headers
+----------------------
+
+Each entry under ``model_providers`` (or the legacy ``llm_providers`` alias) may include a ``headers`` map of extra
+HTTP headers that Plano adds to upstream LLM requests. Plano applies these headers after it sets authentication from
+``access_key`` or ``passthrough_auth``, so you can supply provider-specific metadata without replacing the configured
+credentials.
+
+- **Type:** map of strings (header name → value)
+- **Optional:** yes
+- **Common uses:** required ``User-Agent`` values, organization or account identifiers, or other headers some APIs expect
+
+.. code-block:: yaml
+
+    model_providers:
+      - model: moonshotai/kimi-for-coding
+        access_key: $MOONSHOTAI_API_KEY
+        base_url: https://api.kimi.com/coding/v1
+        headers:
+          User-Agent: "KimiCLI/1.3"
+
+The example below includes this and other provider options in context.
+
 .. literalinclude:: includes/plano_config_full_reference.yaml
     :language: yaml
     :linenos:

docs/source/resources/deployment.rst

@@ -65,7 +65,7 @@ Create a ``docker-compose.yml`` file with the following configuration:
    # docker-compose.yml
    services:
      plano:
-       image: katanemo/plano:0.4.22
+       image: katanemo/plano:0.4.25
        container_name: plano
        ports:
          - "10000:10000" # ingress (client -> plano)
@@ -153,7 +153,7 @@ Create a ``plano-deployment.yaml``:
        spec:
          containers:
            - name: plano
-             image: katanemo/plano:0.4.22
+             image: katanemo/plano:0.4.25
              ports:
                - containerPort: 12000  # LLM gateway (chat completions, model routing)
                  name: llm-gateway

docs/source/resources/includes/plano_config_full_reference.yaml

@@ -47,6 +47,14 @@ model_providers:
     http_host: api.custom-provider.com
     access_key: $CUSTOM_API_KEY
 
+  # headers: optional map of extra HTTP headers sent on upstream requests (after auth).
+  # Use for provider-specific requirements such as User-Agent, org IDs, or account headers.
+  - model: moonshotai/kimi-for-coding
+    access_key: $MOONSHOTAI_API_KEY
+    base_url: https://api.kimi.com/coding/v1
+    headers:
+      User-Agent: "KimiCLI/1.3"
+
 # Model aliases - use friendly names instead of full provider model names
 model_aliases:
   fast-llm:

docs/source/resources/includes/plano_config_full_reference_rendered.yaml

@@ -88,6 +88,18 @@ listeners:
     port: 443
     protocol: https
     provider_interface: openai
+  - access_key: $MOONSHOTAI_API_KEY
+    base_url: https://api.kimi.com/coding/v1
+    base_url_path_prefix: /coding/v1
+    cluster_name: moonshotai_api.kimi.com
+    endpoint: api.kimi.com
+    headers:
+      User-Agent: KimiCLI/1.3
+    model: kimi-for-coding
+    name: moonshotai/kimi-for-coding
+    port: 443
+    protocol: https
+    provider_interface: moonshotai
   name: model_1
   output_filters:
   - input_guards
@@ -144,6 +156,18 @@ model_providers:
   port: 443
   protocol: https
   provider_interface: openai
+- access_key: $MOONSHOTAI_API_KEY
+  base_url: https://api.kimi.com/coding/v1
+  base_url_path_prefix: /coding/v1
+  cluster_name: moonshotai_api.kimi.com
+  endpoint: api.kimi.com
+  headers:
+    User-Agent: KimiCLI/1.3
+  model: kimi-for-coding
+  name: moonshotai/kimi-for-coding
+  port: 443
+  protocol: https
+  provider_interface: moonshotai
 - internal: true
   model: Plano-Orchestrator
   name: plano-orchestrator

skills/AGENTS.md

@@ -31,9 +31,8 @@
   - [5.3 Use `planoai trace` to Inspect Routing Decisions](#use-planoai-trace-to-inspect-routing-decisions)
 - [Section 6: CLI Operations](#section-6)
   - [6.1 Follow the `planoai up` Validation Workflow Before Debugging Runtime Issues](#follow-the-planoai-up-validation-workflow-before-debugging-runtime-issues)
-  - [6.2 Generate Prompt Targets from Python Functions with `planoai generate_prompt_targets`](#generate-prompt-targets-from-python-functions-with-planoai-generateprompttargets)
-  - [6.3 Use `planoai cli_agent` to Connect Claude Code Through Plano](#use-planoai-cliagent-to-connect-claude-code-through-plano)
-  - [6.4 Use `planoai init` Templates to Bootstrap New Projects Correctly](#use-planoai-init-templates-to-bootstrap-new-projects-correctly)
+  - [6.2 Use `planoai cli_agent` to Connect Claude Code Through Plano](#use-planoai-cliagent-to-connect-claude-code-through-plano)
+  - [6.3 Use `planoai init` Templates to Bootstrap New Projects Correctly](#use-planoai-init-templates-to-bootstrap-new-projects-correctly)
 - [Section 7: Deployment & Security](#section-7)
   - [7.1 Understand Plano's Docker Network Topology for Agent URL Configuration](#understand-planos-docker-network-topology-for-agent-url-configuration)
   - [7.2 Use PostgreSQL State Storage for Multi-Turn Conversations in Production](#use-postgresql-state-storage-for-multi-turn-conversations-in-production)
@@ -1377,99 +1376,7 @@ Reference: https://github.com/katanemo/archgw
 
 ---
 
-### 6.2 Generate Prompt Targets from Python Functions with `planoai generate_prompt_targets`
-
-**Impact:** `MEDIUM` — Manually writing prompt_targets YAML for existing Python APIs is error-prone — the generator introspects function signatures and produces correct YAML automatically
-**Tags:** `cli`, `generate`, `prompt-targets`, `python`, `code-generation`
-
-## Generate Prompt Targets from Python Functions with `planoai generate_prompt_targets`
-
-`planoai generate_prompt_targets` introspects Python function signatures and docstrings to generate `prompt_targets` YAML for your Plano config. This is the fastest way to expose existing Python APIs as LLM-callable functions without manually writing the YAML schema.
-
-**Python function requirements for generation:**
-- Use simple type annotations: `int`, `float`, `bool`, `str`, `list`, `tuple`, `set`, `dict`
-- Include a docstring describing what the function does (becomes the `description`)
-- Complex Pydantic models must be flattened into primitive typed parameters first
-
-**Example Python file:**
-
-```python
-# api.py
-
-def get_stock_quote(symbol: str, exchange: str = "NYSE") -> dict:
-    """Get the current stock price and trading data for a given stock symbol.
-
-    Returns price, volume, market cap, and 24h change percentage.
-    """
-    # Implementation calls stock API
-    pass
-
-def get_weather_forecast(city: str, days: int = 3, units: str = "celsius") -> dict:
-    """Get the weather forecast for a city.
-
-    Returns temperature, precipitation, and conditions for the specified number of days.
-    """
-    pass
-
-def search_flights(origin: str, destination: str, date: str, passengers: int = 1) -> list:
-    """Search for available flights between two airports on a given date.
-
-    Date format: YYYY-MM-DD. Returns list of flight options with prices.
-    """
-    pass
-```
-
-**Running the generator:**
-
-```bash
-planoai generate_prompt_targets --file api.py
-```
-
-**Generated output (add to your config.yaml):**
-
-```yaml
-prompt_targets:
-  - name: get_stock_quote
-    description: Get the current stock price and trading data for a given stock symbol.
-    parameters:
-      - name: symbol
-        type: str
-        required: true
-      - name: exchange
-        type: str
-        required: false
-        default: NYSE
-    # Add endpoint manually:
-    endpoint:
-      name: stock_api
-      path: /quote?symbol={symbol}&exchange={exchange}
-
-  - name: get_weather_forecast
-    description: Get the weather forecast for a city.
-    parameters:
-      - name: city
-        type: str
-        required: true
-      - name: days
-        type: int
-        required: false
-        default: 3
-      - name: units
-        type: str
-        required: false
-        default: celsius
-    endpoint:
-      name: weather_api
-      path: /forecast?city={city}&days={days}&units={units}
-```
-
-After generation, manually add the `endpoint` blocks pointing to your actual API. The generator produces the schema; you wire in the connectivity.
-
-Reference: https://github.com/katanemo/archgw
-
----
-
-### 6.3 Use `planoai cli_agent` to Connect Claude Code Through Plano
+### 6.2 Use `planoai cli_agent` to Connect Claude Code Through Plano
 
 **Impact:** `MEDIUM-HIGH` — Running Claude Code directly against provider APIs bypasses Plano's routing, observability, and guardrails — cli_agent routes all Claude Code traffic through your configured Plano instance
 **Tags:** `cli`, `cli-agent`, `claude`, `coding-agent`, `integration`
@@ -1562,7 +1469,7 @@ Reference: [https://github.com/katanemo/archgw](https://github.com/katanemo/arch
 
 ---
 
-### 6.4 Use `planoai init` Templates to Bootstrap New Projects Correctly
+### 6.3 Use `planoai init` Templates to Bootstrap New Projects Correctly
 
 **Impact:** `MEDIUM` — Starting from a blank config.yaml leads to missing required fields and common structural mistakes — templates provide validated, idiomatic starting points
 **Tags:** `cli`, `init`, `templates`, `getting-started`, `project-setup`

skills/README.md

@@ -63,7 +63,7 @@ After installation, these skills are available to your coding agent and can be i
 - `plano-agent-orchestration` - Agent registration and routing descriptions
 - `plano-filter-guardrails` - MCP filters, guardrail messaging, filter ordering
 - `plano-observability-debugging` - Tracing setup, span attributes, trace analysis
-- `plano-cli-operations` - `planoai up`, `cli_agent`, init, prompt target generation
+- `plano-cli-operations` - `planoai up`, `cli_agent`, init
 - `plano-deployment-security` - Docker networking, health checks, state storage
 - `plano-advanced-patterns` - Multi-listener architecture and prompt target schema design
 
@@ -110,7 +110,7 @@ skills/
 | 3 | `agent-` | Agent Orchestration | Descriptions, agent registration |
 | 4 | `filter-` | Filter Chains & Guardrails | Ordering, MCP integration, guardrails |
 | 5 | `observe-` | Observability & Debugging | Tracing, trace inspection, span attributes |
-| 6 | `cli-` | CLI Operations | Startup, CLI agent, init, code generation |
+| 6 | `cli-` | CLI Operations | Startup, CLI agent, init |
 | 7 | `deploy-` | Deployment & Security | Docker networking, state storage, health checks |
 | 8 | `advanced-` | Advanced Patterns | Prompt targets, rate limits, multi-listener |
 

skills/plano-cli-operations/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: plano-cli-operations
-description: Apply Plano CLI best practices. Use for startup troubleshooting, cli_agent workflows, prompt target generation, and template-based project bootstrapping.
+description: Apply Plano CLI best practices. Use for startup troubleshooting, cli_agent workflows, and template-based project bootstrapping.
 license: Apache-2.0
 metadata:
   author: katanemo
@@ -15,20 +15,17 @@ Use this skill when the task is primarily operational and CLI-driven.
 
 - "Fix `planoai up` failures"
 - "Use `planoai cli_agent` with coding agents"
-- "Generate prompt targets from Python functions"
 - "Bootstrap a project with `planoai init` templates"
 
 ## Apply These Rules
 
 - `cli-startup`
 - `cli-agent`
-- `cli-generate`
 - `cli-init`
 
 ## Execution Checklist
 
 1. Follow startup validation order before deep debugging.
 2. Use `cli_agent` to route coding-agent traffic through Plano.
-3. Generate prompt target schema, then wire endpoint details explicitly.
-4. Start from templates for reliable first-time setup.
-5. Provide a compact runbook with exact CLI commands.
+3. Start from templates for reliable first-time setup.
+4. Provide a compact runbook with exact CLI commands.

skills/rules/cli-generate.md

@@ -1,91 +0,0 @@
----
-title: Generate Prompt Targets from Python Functions with `planoai generate_prompt_targets`
-impact: MEDIUM
-impactDescription: Manually writing prompt_targets YAML for existing Python APIs is error-prone — the generator introspects function signatures and produces correct YAML automatically
-tags: cli, generate, prompt-targets, python, code-generation
----
-
-## Generate Prompt Targets from Python Functions with `planoai generate_prompt_targets`
-
-`planoai generate_prompt_targets` introspects Python function signatures and docstrings to generate `prompt_targets` YAML for your Plano config. This is the fastest way to expose existing Python APIs as LLM-callable functions without manually writing the YAML schema.
-
-**Python function requirements for generation:**
-- Use simple type annotations: `int`, `float`, `bool`, `str`, `list`, `tuple`, `set`, `dict`
-- Include a docstring describing what the function does (becomes the `description`)
-- Complex Pydantic models must be flattened into primitive typed parameters first
-
-**Example Python file:**
-
-```python
-# api.py
-
-def get_stock_quote(symbol: str, exchange: str = "NYSE") -> dict:
-    """Get the current stock price and trading data for a given stock symbol.
-
-    Returns price, volume, market cap, and 24h change percentage.
-    """
-    # Implementation calls stock API
-    pass
-
-def get_weather_forecast(city: str, days: int = 3, units: str = "celsius") -> dict:
-    """Get the weather forecast for a city.
-
-    Returns temperature, precipitation, and conditions for the specified number of days.
-    """
-    pass
-
-def search_flights(origin: str, destination: str, date: str, passengers: int = 1) -> list:
-    """Search for available flights between two airports on a given date.
-
-    Date format: YYYY-MM-DD. Returns list of flight options with prices.
-    """
-    pass
-```
-
-**Running the generator:**
-
-```bash
-planoai generate_prompt_targets --file api.py
-```
-
-**Generated output (add to your config.yaml):**
-
-```yaml
-prompt_targets:
-  - name: get_stock_quote
-    description: Get the current stock price and trading data for a given stock symbol.
-    parameters:
-      - name: symbol
-        type: str
-        required: true
-      - name: exchange
-        type: str
-        required: false
-        default: NYSE
-    # Add endpoint manually:
-    endpoint:
-      name: stock_api
-      path: /quote?symbol={symbol}&exchange={exchange}
-
-  - name: get_weather_forecast
-    description: Get the weather forecast for a city.
-    parameters:
-      - name: city
-        type: str
-        required: true
-      - name: days
-        type: int
-        required: false
-        default: 3
-      - name: units
-        type: str
-        required: false
-        default: celsius
-    endpoint:
-      name: weather_api
-      path: /forecast?city={city}&days={days}&units={units}
-```
-
-After generation, manually add the `endpoint` blocks pointing to your actual API. The generator produces the schema; you wire in the connectivity.
-
-Reference: https://github.com/katanemo/archgw