Skip to content

Monocle Support Matrix

This document provides a comprehensive overview of all frameworks and libraries supported by Monocle's instrumentation system. Each framework is instrumented to collect telemetry data including events, attributes, and traces.

Inference Frameworks

Framework Class Method Description
OpenAI Completions create Chat completions, embeddings, and responses. Collects: model, tokens, finish_reason, usage stats
OpenAI AsyncCompletions create Async chat completions. Collects: model, tokens, finish_reason, usage stats
OpenAI Embeddings create Text embeddings generation. Collects: model, input tokens, embedding dimensions
OpenAI AsyncEmbeddings create Async text embeddings generation. Collects: model, input tokens, embedding dimensions
OpenAI Responses create Response generation. Collects: model, tokens, finish_reason, usage stats
OpenAI AsyncResponses create Async response generation. Collects: model, tokens, finish_reason, usage stats
Anthropic Messages create Claude chat completions. Collects: model, input/output tokens, stop_reason
Anthropic AsyncMessages create Async Claude chat completions. Collects: model, input/output tokens, stop_reason
Gemini Models generate_content Google Gemini text generation. Collects: model, tokens, finish_reason, safety ratings
Gemini AsyncModels generate_content Async Google Gemini text generation. Collects: model, tokens, finish_reason, safety ratings
Gemini Models embed_content Google Gemini embeddings. Collects: model, input tokens, embedding dimensions
Mistral Chat complete Mistral chat completions. Collects: model, tokens, finish_reason, usage stats
Mistral Chat complete_async Async Mistral chat completions. Collects: model, tokens, finish_reason, usage stats
Mistral Chat stream Mistral streaming completions. Collects: model, tokens, streaming events
Mistral Chat stream_async Async Mistral streaming completions. Collects: model, tokens, streaming events
Mistral Embeddings create Mistral embeddings generation. Collects: model, input tokens, embedding dimensions
Hugging Face InferenceClient chat_completion Hugging Face chat completions. Collects: model, tokens, finish_reason
Hugging Face AsyncInferenceClient chat_completion Async Hugging Face chat completions. Collects: model, tokens, finish_reason
Azure AI ChatCompletionsClient complete Azure AI chat completions. Collects: model, tokens, finish_reason, usage stats
Azure AI ChatCompletionsClient (async) complete Async Azure AI chat completions. Collects: model, tokens, finish_reason, usage stats
LiteLLM OpenAIChatCompletion completion OpenAI via LiteLLM. Collects: model, tokens, finish_reason, usage stats
LiteLLM AzureChatCompletion completion Azure via LiteLLM. Collects: model, tokens, finish_reason, usage stats
LiteLLM Various providers completion DeepSeek and other providers via LiteLLM. Collects: model, tokens, finish_reason, usage stats. Supports: DeepSeek (api.deepseek.com), Together AI, Perplexity, Groq, Cohere, Fireworks, and more

LangChain

Framework Class Method Description
LangChain BasePromptTemplate invoke Prompt template execution. Collects: template variables, prompt text
LangChain BasePromptTemplate ainvoke Async prompt template execution. Collects: template variables, prompt text
LangChain BaseChatModel invoke Chat model inference. Collects: model, messages, tokens, finish_reason
LangChain BaseChatModel ainvoke Async chat model inference. Collects: model, messages, tokens, finish_reason
LangChain LLM _generate LLM text generation. Collects: model, prompt, tokens, finish_reason
LangChain LLM _agenerate Async LLM text generation. Collects: model, prompt, tokens, finish_reason
LangChain BaseLLM invoke Base LLM invocation. Collects: model, input, tokens, finish_reason
LangChain BaseLLM ainvoke Async base LLM invocation. Collects: model, input, tokens, finish_reason
LangChain BaseRetriever invoke Document retrieval. Collects: query, retrieved documents, scores
LangChain BaseRetriever ainvoke Async document retrieval. Collects: query, retrieved documents, scores
LangChain BaseOutputParser invoke Output parsing. Collects: raw output, parsed result
LangChain BaseOutputParser ainvoke Async output parsing. Collects: raw output, parsed result
LangChain RunnableSequence invoke Chain execution. Collects: input, intermediate outputs, final result
LangChain RunnableSequence ainvoke Async chain execution. Collects: input, intermediate outputs, final result
LangGraph CompiledStateGraph invoke Graph-based agent execution. Collects: agent state, tool calls, decisions
LangGraph CompiledStateGraph ainvoke Async graph-based agent execution. Collects: agent state, tool calls, decisions
LangGraph BaseTool run Tool execution. Collects: tool name, input parameters, output result
LangGraph BaseTool arun Async tool execution. Collects: tool name, input parameters, output result

LlamaIndex

Framework Class Method Description
LlamaIndex BaseRetriever retrieve Document retrieval. Collects: query, retrieved nodes, similarity scores
LlamaIndex BaseRetriever aretrieve Async document retrieval. Collects: query, retrieved nodes, similarity scores
LlamaIndex BaseQueryEngine query Query processing. Collects: query, response, source nodes
LlamaIndex BaseQueryEngine aquery Async query processing. Collects: query, response, source nodes
LlamaIndex CustomLLM chat Custom LLM chat. Collects: model, messages, tokens, finish_reason
LlamaIndex CustomLLM achat Async custom LLM chat. Collects: model, messages, tokens, finish_reason
LlamaIndex OpenAI chat OpenAI LLM via LlamaIndex. Collects: model, messages, tokens, finish_reason
LlamaIndex OpenAI achat Async OpenAI LLM via LlamaIndex. Collects: model, messages, tokens, finish_reason
LlamaIndex MistralAI chat Mistral LLM via LlamaIndex. Collects: model, messages, tokens, finish_reason
LlamaIndex MistralAI achat Async Mistral LLM via LlamaIndex. Collects: model, messages, tokens, finish_reason
LlamaIndex AgentWorkflow run Multi-agent workflow execution. Collects: workflow state, agent interactions
LlamaIndex ReActAgent run ReAct agent execution. Collects: agent reasoning, tool calls, final answer
LlamaIndex ReActAgent achat Async ReAct agent execution. Collects: agent reasoning, tool calls, final answer
LlamaIndex FunctionAgent finalize Function agent finalization. Collects: agent state, final output
LlamaIndex FunctionAgent take_step Function agent step execution. Collects: step input, reasoning, output
LlamaIndex FunctionTool call Function tool execution. Collects: tool name, parameters, result
LlamaIndex FunctionTool acall Async function tool execution. Collects: tool name, parameters, result
LlamaIndex AgentWorkflow _call_tool Multi-agent tool calls. Collects: tool name, parameters, result
LlamaIndex Anthropic chat Anthropic LLM via LlamaIndex. Collects: model, messages, tokens, finish_reason
LlamaIndex Anthropic achat Async Anthropic LLM via LlamaIndex. Collects: model, messages, tokens, finish_reason
LlamaIndex Gemini chat Gemini LLM via LlamaIndex. Collects: model, messages, tokens, finish_reason
LlamaIndex Gemini achat Async Gemini LLM via LlamaIndex. Collects: model, messages, tokens, finish_reason

Haystack

Framework Class Method Description
Haystack InMemoryEmbeddingRetriever run In-memory document retrieval. Collects: query, retrieved documents, scores
Haystack OpenSearchEmbeddingRetriever run OpenSearch document retrieval. Collects: query, retrieved documents, scores
Haystack OpenAIGenerator run OpenAI text generation. Collects: model, prompt, tokens, finish_reason
Haystack OpenAIChatGenerator run OpenAI chat generation. Collects: model, messages, tokens, finish_reason
Haystack Pipeline run Haystack pipeline execution. Collects: pipeline steps, intermediate outputs
Haystack AnthropicChatGenerator run Anthropic chat generation. Collects: model, messages, tokens, finish_reason
Haystack GoogleAIGeminiChatGenerator run Google AI Gemini chat generation. Collects: model, messages, tokens, finish_reason

Web Frameworks

Framework Class Method Description
Flask Flask wsgi_app HTTP request processing. Collects: method, URL, status_code, response_time
Flask Response __call__ HTTP response handling. Collects: status_code, headers, response_size
FastAPI FastAPI __call__ HTTP request processing. Collects: method, URL, status_code, response_time
FastAPI Response __call__ HTTP response handling. Collects: status_code, headers, response_size
aiohttp Application _handle Async HTTP request processing. Collects: method, URL, status_code, response_time
requests Session request HTTP client requests. Collects: method, URL, status_code, response_time

Cloud Platforms

Framework Class Method Description
AWS Botocore ClientCreator create_client AWS service client creation. Collects: service_name, region, client_type
Azure Functions AzureFunctionRouteWrapper run_async Azure Function execution. Collects: function_name, execution_time, status
Azure Functions AzureFunctionRouteWrapper run_sync Azure Function execution. Collects: function_name, execution_time, status
AWS Lambda LambdaFunctionRouteWrapper run_async Lambda Function execution. Collects: function_name, execution_time, status
AWS Lambda LambdaFunctionRouteWrapper run_sync Lambda Function execution. Collects: function_name, execution_time, status

Agent Frameworks

Framework Class Method Description
Google ADK BaseAgent run_async Google ADK agent execution. Collects: agent state, delegation events
Google ADK FunctionTool run_async Google ADK tool execution. Collects: tool name, parameters, result
Google ADK Runner run_async Google ADK runner execution. Collects: request context, execution flow
Agents Runner run Agent runner execution. Collects: agent state, tool calls, decisions
Agents Runner run_sync Sync agent runner execution. Collects: agent state, tool calls, decisions
Agents AgentRunner _run_single_turn Single agent turn execution. Collects: turn input, reasoning, output
Agents FunctionTool __init__ Function tool initialization. Collects: tool definition, parameters
Agents Handoff __init__ Agent handoff initialization. Collects: handoff context, delegation info
Teams AI OpenAIModel complete_prompt Teams AI prompt completion. Collects: model, prompt, response, tokens
Teams AI ActionPlanner complete_prompt Teams AI action planning. Collects: plan, actions, reasoning
A2A A2AClient send_message A2A client messaging. Collects: message content, routing info, response

Protocol Frameworks

Framework Class Method Description
MCP BaseSession send_request MCP protocol requests. Collects: request type, parameters, response
MCP convert_mcp_tool_to_langchain_tool N/A MCP to LangChain tool conversion. Collects: tool definition, conversion result

Provider URLs

The following provider URLs are configured or extracted by Monocle's instrumentation:

Provider Default API Endpoint Notes
OpenAI https://api.openai.com/v1/ Extracted from client configuration
Anthropic https://api.anthropic.com Extracted from client configuration
Mistral https://api.mistral.ai Hardcoded default endpoint
Gemini https://generativelanguage.googleapis.com/ Extracted from client configuration
Hugging Face https://api-inference.huggingface.co/v1/ Hardcoded default endpoint
Azure OpenAI Variable (Azure endpoint) Extracted from client configuration
Azure AI Inference Variable (Azure endpoint) Extracted from client configuration
DeepSeek https://api.deepseek.com Supported via LiteLLM
LiteLLM Providers Variable Supports multiple providers including DeepSeek, Together AI, Perplexity, Groq, Cohere, Fireworks, and more. Endpoint extracted from client configuration

Data Collection Summary

Each framework collects relevant telemetry data including:

  • Inference Events: Model calls, token usage, finish reasons, response times
  • Retrieval Events: Query processing, document retrieval, similarity scores
  • Agent Events: Agent reasoning, tool calls, delegation events, workflow state
  • HTTP Events: Request/response data, status codes, response times, headers
  • Tool Events: Tool execution, parameters, results, error handling
  • Pipeline Events: Multi-step execution, intermediate outputs, final results

This comprehensive instrumentation enables detailed observability across the entire AI application stack, from individual model calls to complex multi-agent workflows.