Merge pull request #313 from trustgraph-ai/feature/mistral-support

Added support for Mistral API

Containerfile

@@ -15,7 +15,7 @@ RUN pip3 install torch==2.5.1+cpu \
     --index-url https://download.pytorch.org/whl/cpu
 
 RUN pip3 install \
-    anthropic boto3 cohere openai google-cloud-aiplatform \
+    anthropic boto3 cohere mistralai openai google-cloud-aiplatform \
     ollama google-generativeai \
     langchain==0.3.13 langchain-core==0.3.28 langchain-huggingface==0.1.2 \
     langchain-text-splitters==0.3.4 \

Makefile

@@ -68,13 +68,13 @@ clean:
 set-version:
 	echo '"${VERSION}"' > templates/values/version.jsonnet
 
-TEMPLATES=azure bedrock claude cohere mix llamafile ollama openai vertexai \
+TEMPLATES=azure bedrock claude cohere mix llamafile mistral ollama openai vertexai \
     openai-neo4j storage
 
 DCS=$(foreach template,${TEMPLATES},${template:%=tg-launch-%.yaml})
 
-MODELS=azure bedrock claude cohere llamafile ollama openai vertexai
-GRAPHS=cassandra neo4j falkordb
+MODELS=azure bedrock claude cohere llamafile mistral ollama openai vertexai
+GRAPHS=cassandra neo4j falkordb memgraph
 
 # tg-launch-%.yaml: templates/%.jsonnet templates/components/version.jsonnet
 # 	jsonnet -Jtemplates \

containers/Containerfile.flow

@@ -12,7 +12,7 @@ RUN dnf install -y python3 python3-pip python3-wheel python3-aiohttp \
     python3-rdflib
 
 RUN pip3 install --no-cache-dir \
-    anthropic cohere openai google-generativeai \
+    anthropic cohere mistralai openai google-generativeai \
     ollama \
     langchain==0.3.13 langchain-core==0.3.28 \
     langchain-text-splitters==0.3.4 \

templates/all-patterns.jsonnet

@@ -13,6 +13,7 @@
     import "patterns/llm-claude.jsonnet",
     import "patterns/llm-cohere.jsonnet",
     import "patterns/llm-llamafile.jsonnet",
+    import "patterns/llm-mistral.jsonnet",
     import "patterns/llm-ollama.jsonnet",
     import "patterns/llm-openai.jsonnet",
     import "patterns/llm-vertexai.jsonnet",

templates/components.jsonnet

@@ -11,6 +11,7 @@
    "claude": import "components/claude.jsonnet",
    "cohere": import "components/cohere.jsonnet",
    "googleaistudio": import "components/googleaistudio.jsonnet",
+   "mistral": import "components/mistral.jsonnet",
    "ollama": import "components/ollama.jsonnet",
    "openai": import "components/openai.jsonnet",
    "vertexai": import "components/vertexai.jsonnet",
@@ -22,6 +23,7 @@
    "claude-rag": import "components/claude-rag.jsonnet",
    "cohere-rag": import "components/cohere-rag.jsonnet",
    "googleaistudio-rag": import "components/googleaistudio-rag.jsonnet",
+   "mistral-rag": import "components/mistral-rag.jsonnet",
    "ollama-rag": import "components/ollama-rag.jsonnet",
    "openai-rag": import "components/openai-rag.jsonnet",
    "vertexai-rag": import "components/vertexai-rag.jsonnet",

templates/components/mistral-rag.jsonnet

@@ -0,0 +1,63 @@
+local base = import "base/base.jsonnet";
+local images = import "values/images.jsonnet";
+local url = import "values/url.jsonnet";
+local prompts = import "prompts/mixtral.jsonnet";
+
+{
+
+    with:: function(key, value)
+        self + {
+            ["mistral-rag-" + key]:: value,
+        },
+
+    "mistral-rag-max-output-tokens":: 4096,
+    "mistral-rag-temperature":: 0.0,
+    "mistral-rag-model":: "ministral-8b-latest",
+
+    "text-completion-rag" +: {
+    
+        create:: function(engine)
+
+            local envSecrets = engine.envSecrets("mistral-credentials")
+                .with_env_var("MISTRAL_TOKEN", "mistral-token");
+
+            local containerRag =
+                engine.container("text-completion-rag")
+                    .with_image(images.trustgraph_flow)
+                    .with_command([
+                        "text-completion-mistral",
+                        "-p",
+                        url.pulsar,
+                        "-x",
+                        std.toString($["mistral-rag-max-output-tokens"]),
+                        "-t",
+                        "%0.3f" % $["mistral-rag-temperature"],
+                        "-m",
+                        $["mistral-rag-model"],
+                        "-i",
+                        "non-persistent://tg/request/text-completion-rag",
+                        "-o",
+                        "non-persistent://tg/response/text-completion-rag",
+                    ])
+                    .with_env_var_secrets(envSecrets)
+                    .with_limits("0.5", "128M")
+                    .with_reservations("0.1", "128M");
+
+            local containerSetRag = engine.containers(
+                "text-completion-rag", [ containerRag ]
+            );
+
+            local serviceRag =
+                engine.internalService(containerSetRag)
+                .with_port(8080, 8080, "metrics");
+
+            engine.resources([
+                envSecrets,
+                containerSetRag,
+                serviceRag,
+            ])
+
+    },
+
+} + prompts
+

templates/components/mistral.jsonnet

@@ -0,0 +1,59 @@
+local base = import "base/base.jsonnet";
+local images = import "values/images.jsonnet";
+local url = import "values/url.jsonnet";
+local prompts = import "prompts/mixtral.jsonnet";
+
+{
+
+    with:: function(key, value)
+        self + {
+            ["mistral-" + key]:: value,
+        },
+
+    "mistral-max-output-tokens":: 4096,
+    "mistral-temperature":: 0.0,
+    "mistral-model":: "ministral-8b-latest",
+
+    "text-completion" +: {
+    
+        create:: function(engine)
+
+            local envSecrets = engine.envSecrets("mistral-credentials")
+                .with_env_var("MISTRAL_TOKEN", "mistral-token");
+
+            local container =
+                engine.container("text-completion")
+                    .with_image(images.trustgraph_flow)
+                    .with_command([
+                        "text-completion-mistral",
+                        "-p",
+                        url.pulsar,
+                        "-x",
+                        std.toString($["mistral-max-output-tokens"]),
+                        "-t",
+                        "%0.3f" % $["mistral-temperature"],
+                        "-m",
+                        $["mistral-model"],
+                    ])
+                    .with_env_var_secrets(envSecrets)
+                    .with_limits("0.5", "128M")
+                    .with_reservations("0.1", "128M");
+
+            local containerSet = engine.containers(
+                "text-completion", [ container ]
+            );
+
+            local service =
+                engine.internalService(containerSet)
+                .with_port(8080, 8080, "metrics");
+
+            engine.resources([
+                envSecrets,
+                containerSet,
+                service,
+            ])
+
+    },
+
+} + prompts
+

templates/generate-all

@@ -134,7 +134,7 @@ def generate_all(output, version):
     ]:
         for model in [
                 # "azure", "azure-openai", "bedrock", "claude", "cohere",
-                # "googleaistudio", "llamafile",
+                # "googleaistudio", "llamafile", "mistral",
                 "ollama",
                 # "openai", "vertexai",
         ]:

templates/patterns/llm-mistral.jsonnet

@@ -0,0 +1,32 @@
+{
+    pattern: {
+	name: "mistral",
+        icon: "🤖💬",
+        title: "Add Mistral LLM endpoint for text completion",
+	description: "This pattern integrates a Mistral LLM service for text completion operations.  You need a Mistral subscription and have an API key to be able to use this service.",
+        requires: ["pulsar", "trustgraph"],
+        features: ["llm"],
+	args: [
+	    {
+		name: "mistral-max-output-tokens",
+                label: "Maximum output tokens",
+		type: "integer",
+		description: "Limit on number tokens to generate",
+                default: 4096,
+		required: true,
+            },
+	    {
+		name: "mistral-temperature",
+                label: "Temperature",
+		type: "slider",
+		description: "Controlling predictability / creativity balance",
+                min: 0,
+                max: 1,
+                step: 0.05,
+                default: 0.5,
+            },
+	],
+        category: [ "llm" ],
+    },
+    module: "components/mistral.jsonnet",
+}

trustgraph-flow/scripts/text-completion-mistral

@@ -0,0 +1,6 @@
+#!/usr/bin/env python3
+
+from trustgraph.model.text_completion.mistral import run
+
+run()
+

trustgraph-flow/setup.py

@@ -50,6 +50,7 @@ setuptools.setup(
         "langchain-core",
         "langchain-text-splitters",
         "minio",
+        "mistralai"
         "neo4j",
         "ollama",
         "openai",
@@ -107,6 +108,7 @@ setuptools.setup(
         "scripts/text-completion-cohere",
         "scripts/text-completion-googleaistudio",
         "scripts/text-completion-llamafile",
+        "scripts/text-completion-mistral",
         "scripts/text-completion-ollama",
         "scripts/text-completion-openai",
         "scripts/triples-query-cassandra",

trustgraph-flow/trustgraph/model/text_completion/mistral/__init__.py

@@ -0,0 +1,3 @@
+
+from . llm import *
+

trustgraph-flow/trustgraph/model/text_completion/mistral/__main__.py

@@ -0,0 +1,7 @@
+#!/usr/bin/env python3
+
+from . llm import run
+
+if __name__ == '__main__':
+    run()
+

trustgraph-flow/trustgraph/model/text_completion/mistral/llm.py

@@ -0,0 +1,201 @@
+
+"""
+Simple LLM service, performs text prompt completion using Mistral.
+Input is prompt, output is response.
+"""
+
+from mistralai import Mistral, RateLimitError
+from prometheus_client import Histogram
+import os
+
+from .... schema import TextCompletionRequest, TextCompletionResponse, Error
+from .... schema import text_completion_request_queue
+from .... schema import text_completion_response_queue
+from .... log_level import LogLevel
+from .... base import ConsumerProducer
+from .... exceptions import TooManyRequests
+
+module = ".".join(__name__.split(".")[1:-1])
+
+default_input_queue = text_completion_request_queue
+default_output_queue = text_completion_response_queue
+default_subscriber = module
+default_model = 'ministral-8b-latest'
+default_temperature = 0.0
+default_max_output = 4096
+default_api_key = os.getenv("MISTRAL_TOKEN")
+
+class Processor(ConsumerProducer):
+
+    def __init__(self, **params):
+    
+        input_queue = params.get("input_queue", default_input_queue)
+        output_queue = params.get("output_queue", default_output_queue)
+        subscriber = params.get("subscriber", default_subscriber)
+        model = params.get("model", default_model)
+        api_key = params.get("api_key", default_api_key)
+        temperature = params.get("temperature", default_temperature)
+        max_output = params.get("max_output", default_max_output)
+
+        if api_key is None:
+            raise RuntimeError("Mistral API key not specified")
+
+        super(Processor, self).__init__(
+            **params | {
+                "input_queue": input_queue,
+                "output_queue": output_queue,
+                "subscriber": subscriber,
+                "input_schema": TextCompletionRequest,
+                "output_schema": TextCompletionResponse,
+                "model": model,
+                "temperature": temperature,
+                "max_output": max_output,
+            }
+        )
+
+        if not hasattr(__class__, "text_completion_metric"):
+            __class__.text_completion_metric = Histogram(
+                'text_completion_duration',
+                'Text completion duration (seconds)',
+                buckets=[
+                    0.25, 0.5, 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0,
+                    8.0, 9.0, 10.0, 11.0, 12.0, 13.0, 14.0, 15.0, 16.0,
+                    17.0, 18.0, 19.0, 20.0, 21.0, 22.0, 23.0, 24.0, 25.0,
+                    30.0, 35.0, 40.0, 45.0, 50.0, 60.0, 80.0, 100.0,
+                    120.0
+                ]
+            )
+
+        self.model = model
+        self.temperature = temperature
+        self.max_output = max_output
+        self.mistral = Mistral(api_key=api_key)
+
+        print("Initialised", flush=True)
+
+    async def handle(self, msg):
+
+        v = msg.value()
+
+        # Sender-produced ID
+
+        id = msg.properties()["id"]
+
+        print(f"Handling prompt {id}...", flush=True)
+
+        prompt = v.system + "\n\n" + v.prompt
+
+        try:
+
+            with __class__.text_completion_metric.time():
+
+                resp = self.mistral.chat.complete(
+                    model=self.model,
+                    messages=[
+                        {
+                            "role": "user",
+                            "content": [
+                                {
+                                    "type": "text",
+                                    "text": prompt
+                                }
+                            ]
+                        }
+                    ],
+                    temperature=self.temperature,
+                    max_tokens=self.max_output,
+                    top_p=1,
+                    frequency_penalty=0,
+                    presence_penalty=0,
+                    response_format={
+                        "type": "text"
+                    }
+                )
+            
+            inputtokens = resp.usage.prompt_tokens
+            outputtokens = resp.usage.completion_tokens
+            print(resp.choices[0].message.content, flush=True)
+            print(f"Input Tokens: {inputtokens}", flush=True)
+            print(f"Output Tokens: {outputtokens}", flush=True)
+
+            print("Send response...", flush=True)
+            r = TextCompletionResponse(
+                response=resp.choices[0].message.content,
+                error=None,
+                in_token=inputtokens,
+                out_token=outputtokens,
+                model=self.model
+            )
+            await self.send(r, properties={"id": id})
+
+            print("Done.", flush=True)
+
+        # FIXME: Wrong exception, don't know what this LLM throws
+        # for a rate limit
+        except Mistral.RateLimitError:
+
+            # Leave rate limit retries to the base handler
+            raise TooManyRequests()
+
+        except Exception as e:
+
+            # Apart from rate limits, treat all exceptions as unrecoverable
+
+            print(f"Exception: {e}")
+
+            print("Send error response...", flush=True)
+
+            r = TextCompletionResponse(
+                error=Error(
+                    type = "llm-error",
+                    message = str(e),
+                ),
+                response=None,
+                in_token=None,
+                out_token=None,
+                model=None,
+            )
+
+            await self.send(r, properties={"id": id})
+
+            self.consumer.acknowledge(msg)
+
+    @staticmethod
+    def add_args(parser):
+
+        ConsumerProducer.add_args(
+            parser, default_input_queue, default_subscriber,
+            default_output_queue,
+        )
+
+        parser.add_argument(
+            '-m', '--model',
+            default=default_model,
+            help=f'LLM model (default: ministral-8b-latest)'
+        )
+
+        parser.add_argument(
+            '-k', '--api-key',
+            default=default_api_key,
+            help=f'Mistral API Key'
+        )
+
+        parser.add_argument(
+            '-t', '--temperature',
+            type=float,
+            default=default_temperature,
+            help=f'LLM temperature parameter (default: {default_temperature})'
+        )
+
+        parser.add_argument(
+            '-x', '--max-output',
+            type=int,
+            default=default_max_output,
+            help=f'LLM max output tokens (default: {default_max_output})'
+        )
+
+def run():
+
+    Processor.launch(module, __doc__)
+
+