rowboat/apps/x/packages/shared/src/models.ts

61 lines
2.9 KiB
TypeScript
Raw Normal View History

import { z } from "zod";
export const LlmProvider = z.object({
freeze model + provider per run at creation time The model dropdown was broken in two ways: it wrote to ~/.rowboat/config/models.json (the BYOK creds file, stamped with a fake `flavor: 'openrouter'` to satisfy zod when signed in), and the runtime ignored that write entirely for signed-in users because `streamAgent` hard-coded `gpt-5.4`. Model selection was also globally scoped, so every chat shared one brain. This change moves model + provider out of the global config and onto the run itself, resolved once at runs:create and frozen for the run's lifetime. ## Resolution `runsCore.createRun` resolves per-field, falling through: run.model = opts.model ?? agent.model ?? defaults.model run.provider = opts.provider ?? agent.provider ?? defaults.provider A new `core/models/defaults.ts` is the only place in the codebase that branches on signed-in state. `getDefaultModelAndProvider()` returns name strings; `resolveProviderConfig(name)` does the name → full LlmProvider lookup at runtime. `createProvider` learns about `flavor: 'rowboat'` so the gateway is just another flavor. `provider` is stored as a name (e.g. `"rowboat"`, `"openai"`), not a full LlmProvider object. API keys never get written into the JSONL log; rotating a key in models.json applies to existing runs without re-creation. Cost: deleting a provider from settings breaks runs that referenced it (clear error surfaced via `resolveProviderConfig`). ## Runtime `streamAgent` no longer resolves anything — it reads `state.runModel` / `state.runProvider`, looks up the provider config, instantiates. Subflows inherit the parent run's pair, so KG / inline-task subagents run on whatever the main run resolved to at creation. The `knowledgeGraphAgents` array, `isKgAgent`, and the per-agent default constants are gone. KG / inline-task / pre-built agents declare their preferred model in YAML frontmatter (claude-haiku-4.5 / claude-sonnet-4.6) — used at resolution time when those agents are themselves the top-level agent of a run (background triggers, scheduled tasks, etc.). ## Standalone callers Non-run LLM call sites (summarize_meeting, track/routing, builtin-tools parseFile) and `agent-schedule/runner` were branching on signed-in independently. They all route through `getDefaultModelAndProvider` + `resolveProviderConfig` + `createProvider` now; `agent-schedule/runner` switched from raw `runsRepo.create` to `runsCore.createRun` so resolution applies to scheduled-agent runs too. ## UI `chat-input-with-mentions` stops calling `models:saveConfig`. The dropdown notifies the parent via `onSelectedModelChange` ({provider, model} as names); App.tsx stashes selection per-tab and passes it to the next `runs:create`. When a run already exists, the input fetches it and renders a static label — model can't change mid-run. ## Legacy runs A lenient zod schema in `repo.ts` (`StartEvent.extend(...optional)` plus `RunEvent.or(LegacyStartEvent)`) parses pre-existing runs. `repo.fetch` fills missing model/provider from current defaults and returns the strict canonical `Run` type. No file-rewriting migration; no impact on the canonical schema in `@x/shared`. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-22 12:26:01 +05:30
flavor: z.enum(["openai", "anthropic", "google", "openrouter", "aigateway", "ollama", "openai-compatible", "rowboat"]),
apiKey: z.string().optional(),
baseURL: z.string().optional(),
headers: z.record(z.string(), z.string()).optional(),
2026-07-05 11:45:22 +05:30
// Context window (in tokens) to request from local runtimes. Ollama defaults
// to a ~4k window that silently truncates Rowboat's prompts; when unset,
// local providers get a larger default (see core/models/local.ts).
contextLength: z.number().int().positive().optional(),
2026-07-05 14:50:15 +05:30
// Reasoning effort for local thinking models (Ollama `think` parameter).
// gpt-oss supports the levels directly; other thinking models map
// low → thinking off, high → thinking on. Defaults to "low" for Ollama —
// background agents and chat both want snappy responses on local hardware.
reasoningEffort: z.enum(["low", "medium", "high"]).optional(),
});
2026-07-05 16:46:32 +05:30
// A provider-qualified model reference. `provider` is a provider name as
// understood by resolveProviderConfig — a BYOK flavor ("ollama", "openai",
// …) or "rowboat" for the signed-in gateway.
export const ModelRef = z.object({
provider: z.string(),
model: z.string(),
});
// Category overrides accept either a bare model id (legacy: paired with the
// active default provider) or a provider-qualified ref (hybrid mode: e.g.
// gateway assistant + local Ollama background agents).
export const ModelOverride = z.union([z.string(), ModelRef]);
export const LlmModelConfig = z.object({
provider: LlmProvider,
model: z.string(),
models: z.array(z.string()).optional(),
2026-07-05 16:46:32 +05:30
// The user's explicit default assistant model. When set it wins over both
// the signed-in curated default and the legacy top-level provider/model
// pair — this is what lets signed-in users default to a BYOK model.
defaultSelection: ModelRef.optional(),
freeze model + provider per run at creation time The model dropdown was broken in two ways: it wrote to ~/.rowboat/config/models.json (the BYOK creds file, stamped with a fake `flavor: 'openrouter'` to satisfy zod when signed in), and the runtime ignored that write entirely for signed-in users because `streamAgent` hard-coded `gpt-5.4`. Model selection was also globally scoped, so every chat shared one brain. This change moves model + provider out of the global config and onto the run itself, resolved once at runs:create and frozen for the run's lifetime. ## Resolution `runsCore.createRun` resolves per-field, falling through: run.model = opts.model ?? agent.model ?? defaults.model run.provider = opts.provider ?? agent.provider ?? defaults.provider A new `core/models/defaults.ts` is the only place in the codebase that branches on signed-in state. `getDefaultModelAndProvider()` returns name strings; `resolveProviderConfig(name)` does the name → full LlmProvider lookup at runtime. `createProvider` learns about `flavor: 'rowboat'` so the gateway is just another flavor. `provider` is stored as a name (e.g. `"rowboat"`, `"openai"`), not a full LlmProvider object. API keys never get written into the JSONL log; rotating a key in models.json applies to existing runs without re-creation. Cost: deleting a provider from settings breaks runs that referenced it (clear error surfaced via `resolveProviderConfig`). ## Runtime `streamAgent` no longer resolves anything — it reads `state.runModel` / `state.runProvider`, looks up the provider config, instantiates. Subflows inherit the parent run's pair, so KG / inline-task subagents run on whatever the main run resolved to at creation. The `knowledgeGraphAgents` array, `isKgAgent`, and the per-agent default constants are gone. KG / inline-task / pre-built agents declare their preferred model in YAML frontmatter (claude-haiku-4.5 / claude-sonnet-4.6) — used at resolution time when those agents are themselves the top-level agent of a run (background triggers, scheduled tasks, etc.). ## Standalone callers Non-run LLM call sites (summarize_meeting, track/routing, builtin-tools parseFile) and `agent-schedule/runner` were branching on signed-in independently. They all route through `getDefaultModelAndProvider` + `resolveProviderConfig` + `createProvider` now; `agent-schedule/runner` switched from raw `runsRepo.create` to `runsCore.createRun` so resolution applies to scheduled-agent runs too. ## UI `chat-input-with-mentions` stops calling `models:saveConfig`. The dropdown notifies the parent via `onSelectedModelChange` ({provider, model} as names); App.tsx stashes selection per-tab and passes it to the next `runs:create`. When a run already exists, the input fetches it and renders a static label — model can't change mid-run. ## Legacy runs A lenient zod schema in `repo.ts` (`StartEvent.extend(...optional)` plus `RunEvent.or(LegacyStartEvent)`) parses pre-existing runs. `repo.fetch` fills missing model/provider from current defaults and returns the strict canonical `Run` type. No file-rewriting migration; no impact on the canonical schema in `@x/shared`. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-22 12:26:01 +05:30
providers: z.record(z.string(), z.object({
apiKey: z.string().optional(),
baseURL: z.string().optional(),
headers: z.record(z.string(), z.string()).optional(),
2026-07-05 11:45:22 +05:30
contextLength: z.number().int().positive().optional(),
2026-07-05 14:50:15 +05:30
reasoningEffort: z.enum(["low", "medium", "high"]).optional(),
freeze model + provider per run at creation time The model dropdown was broken in two ways: it wrote to ~/.rowboat/config/models.json (the BYOK creds file, stamped with a fake `flavor: 'openrouter'` to satisfy zod when signed in), and the runtime ignored that write entirely for signed-in users because `streamAgent` hard-coded `gpt-5.4`. Model selection was also globally scoped, so every chat shared one brain. This change moves model + provider out of the global config and onto the run itself, resolved once at runs:create and frozen for the run's lifetime. ## Resolution `runsCore.createRun` resolves per-field, falling through: run.model = opts.model ?? agent.model ?? defaults.model run.provider = opts.provider ?? agent.provider ?? defaults.provider A new `core/models/defaults.ts` is the only place in the codebase that branches on signed-in state. `getDefaultModelAndProvider()` returns name strings; `resolveProviderConfig(name)` does the name → full LlmProvider lookup at runtime. `createProvider` learns about `flavor: 'rowboat'` so the gateway is just another flavor. `provider` is stored as a name (e.g. `"rowboat"`, `"openai"`), not a full LlmProvider object. API keys never get written into the JSONL log; rotating a key in models.json applies to existing runs without re-creation. Cost: deleting a provider from settings breaks runs that referenced it (clear error surfaced via `resolveProviderConfig`). ## Runtime `streamAgent` no longer resolves anything — it reads `state.runModel` / `state.runProvider`, looks up the provider config, instantiates. Subflows inherit the parent run's pair, so KG / inline-task subagents run on whatever the main run resolved to at creation. The `knowledgeGraphAgents` array, `isKgAgent`, and the per-agent default constants are gone. KG / inline-task / pre-built agents declare their preferred model in YAML frontmatter (claude-haiku-4.5 / claude-sonnet-4.6) — used at resolution time when those agents are themselves the top-level agent of a run (background triggers, scheduled tasks, etc.). ## Standalone callers Non-run LLM call sites (summarize_meeting, track/routing, builtin-tools parseFile) and `agent-schedule/runner` were branching on signed-in independently. They all route through `getDefaultModelAndProvider` + `resolveProviderConfig` + `createProvider` now; `agent-schedule/runner` switched from raw `runsRepo.create` to `runsCore.createRun` so resolution applies to scheduled-agent runs too. ## UI `chat-input-with-mentions` stops calling `models:saveConfig`. The dropdown notifies the parent via `onSelectedModelChange` ({provider, model} as names); App.tsx stashes selection per-tab and passes it to the next `runs:create`. When a run already exists, the input fetches it and renders a static label — model can't change mid-run. ## Legacy runs A lenient zod schema in `repo.ts` (`StartEvent.extend(...optional)` plus `RunEvent.or(LegacyStartEvent)`) parses pre-existing runs. `repo.fetch` fills missing model/provider from current defaults and returns the strict canonical `Run` type. No file-rewriting migration; no impact on the canonical schema in `@x/shared`. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-22 12:26:01 +05:30
model: z.string().optional(),
models: z.array(z.string()).optional(),
knowledgeGraphModel: z.string().optional(),
meetingNotesModel: z.string().optional(),
liveNoteAgentModel: z.string().optional(),
autoPermissionDecisionModel: z.string().optional(),
freeze model + provider per run at creation time The model dropdown was broken in two ways: it wrote to ~/.rowboat/config/models.json (the BYOK creds file, stamped with a fake `flavor: 'openrouter'` to satisfy zod when signed in), and the runtime ignored that write entirely for signed-in users because `streamAgent` hard-coded `gpt-5.4`. Model selection was also globally scoped, so every chat shared one brain. This change moves model + provider out of the global config and onto the run itself, resolved once at runs:create and frozen for the run's lifetime. ## Resolution `runsCore.createRun` resolves per-field, falling through: run.model = opts.model ?? agent.model ?? defaults.model run.provider = opts.provider ?? agent.provider ?? defaults.provider A new `core/models/defaults.ts` is the only place in the codebase that branches on signed-in state. `getDefaultModelAndProvider()` returns name strings; `resolveProviderConfig(name)` does the name → full LlmProvider lookup at runtime. `createProvider` learns about `flavor: 'rowboat'` so the gateway is just another flavor. `provider` is stored as a name (e.g. `"rowboat"`, `"openai"`), not a full LlmProvider object. API keys never get written into the JSONL log; rotating a key in models.json applies to existing runs without re-creation. Cost: deleting a provider from settings breaks runs that referenced it (clear error surfaced via `resolveProviderConfig`). ## Runtime `streamAgent` no longer resolves anything — it reads `state.runModel` / `state.runProvider`, looks up the provider config, instantiates. Subflows inherit the parent run's pair, so KG / inline-task subagents run on whatever the main run resolved to at creation. The `knowledgeGraphAgents` array, `isKgAgent`, and the per-agent default constants are gone. KG / inline-task / pre-built agents declare their preferred model in YAML frontmatter (claude-haiku-4.5 / claude-sonnet-4.6) — used at resolution time when those agents are themselves the top-level agent of a run (background triggers, scheduled tasks, etc.). ## Standalone callers Non-run LLM call sites (summarize_meeting, track/routing, builtin-tools parseFile) and `agent-schedule/runner` were branching on signed-in independently. They all route through `getDefaultModelAndProvider` + `resolveProviderConfig` + `createProvider` now; `agent-schedule/runner` switched from raw `runsRepo.create` to `runsCore.createRun` so resolution applies to scheduled-agent runs too. ## UI `chat-input-with-mentions` stops calling `models:saveConfig`. The dropdown notifies the parent via `onSelectedModelChange` ({provider, model} as names); App.tsx stashes selection per-tab and passes it to the next `runs:create`. When a run already exists, the input fetches it and renders a static label — model can't change mid-run. ## Legacy runs A lenient zod schema in `repo.ts` (`StartEvent.extend(...optional)` plus `RunEvent.or(LegacyStartEvent)`) parses pre-existing runs. `repo.fetch` fills missing model/provider from current defaults and returns the strict canonical `Run` type. No file-rewriting migration; no impact on the canonical schema in `@x/shared`. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-22 12:26:01 +05:30
})).optional(),
2026-07-05 16:46:32 +05:30
// Per-category model overrides. Honored in both modes: when unset,
// signed-in users get the curated gateway defaults and BYOK users get the
// assistant model. Read by helpers in core/models/defaults.ts.
knowledgeGraphModel: ModelOverride.optional(),
meetingNotesModel: ModelOverride.optional(),
liveNoteAgentModel: ModelOverride.optional(),
autoPermissionDecisionModel: ModelOverride.optional(),
});