Added basic Mistral API support

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 \

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__)
+
+