Skip to main content
LiteLLM provides a unified interface for calling 100+ LLMs using the OpenAI format. HoneyHive integrates with LiteLLM via the OpenInference instrumentor, automatically capturing all completion calls, token usage, and model metadata across providers.

​
Quick Start

Add HoneyHive tracing in just 4 lines of code. Add this to your existing LiteLLM app and all completion calls are automatically traced, regardless of the underlying provider.
To see where to initialize the tracer for your environment, including AWS Lambda and long-running servers, see Tracer Initialization.
pip install "honeyhive[openinference-litellm]>=1.0.0rc0"

# Or install separately
pip install "honeyhive>=1.0.0rc0" openinference-instrumentation-litellm litellm

import os
import litellm
from honeyhive import HoneyHiveTracer
from openinference.instrumentation.litellm import LiteLLMInstrumentor

tracer = HoneyHiveTracer.init(
    api_key=os.getenv("HH_API_KEY"),
    project=os.getenv("HH_PROJECT"),
)
LiteLLMInstrumentor().instrument(tracer_provider=tracer.provider)

# Your existing LiteLLM code works unchanged

​
What Gets Traced

The instrumentor automatically captures:
  • Completions - litellm.completion() and litellm.acompletion() with inputs, outputs, and token usage
  • Embeddings - litellm.embedding() and litellm.aembedding() requests
  • Image generation - litellm.image_generation() and litellm.aimage_generation() calls
  • Multi-provider routing - Model name and provider metadata for every call
No manual instrumentation required.
Use module-level calls. The instrumentor patches litellm.completion(), not directly imported functions. Always use litellm.completion() instead of from litellm import completion.

​
Example: Multi-Provider Completions

import os
import litellm
from honeyhive import HoneyHiveTracer
from openinference.instrumentation.litellm import LiteLLMInstrumentor

tracer = HoneyHiveTracer.init(
    api_key=os.getenv("HH_API_KEY"),
    project=os.getenv("HH_PROJECT"),
)
LiteLLMInstrumentor().instrument(tracer_provider=tracer.provider)

# OpenAI via LiteLLM
response = litellm.completion(
    model="openai/gpt-4o-mini",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "What is the capital of France?"},
    ],
    max_tokens=50,
)
print(response.choices[0].message.content)

# Anthropic via LiteLLM - same interface, automatically traced
response2 = litellm.completion(
    model="anthropic/claude-haiku-4-5-20251001",
    messages=[
        {"role": "user", "content": "Tell me a fun fact about Paris."},
    ],
    max_tokens=100,
)
print(response2.choices[0].message.content)

​
Avoiding Duplicate Spans

Avoid duplicate spans. If you also use provider-specific instrumentors (e.g., OpenAIInstrumentor), LiteLLM calls to that provider may produce duplicate spans. Use only the LiteLLM instrumentor when routing calls through LiteLLM.

​
Environment Configuration

# HoneyHive configuration
export HH_API_KEY="your-honeyhive-api-key"
export HH_PROJECT="your-project"

# Provider API keys (set whichever providers you use)
export OPENAI_API_KEY="your-openai-api-key"
export ANTHROPIC_API_KEY="your-anthropic-api-key"

​
Troubleshooting

​
Traces not appearing

  1. Check environment variables - Ensure HH_API_KEY and HH_PROJECT are set
  2. Pass the tracer provider - The instrumentor must receive tracer_provider=tracer.provider:
import os
from honeyhive import HoneyHiveTracer
from openinference.instrumentation.litellm import LiteLLMInstrumentor

tracer = HoneyHiveTracer.init(
    api_key=os.getenv("HH_API_KEY"),
    project=os.getenv("HH_PROJECT"),
)

# ✅ Correct - pass tracer_provider
LiteLLMInstrumentor().instrument(tracer_provider=tracer.provider)

# ❌ Wrong - missing tracer_provider
LiteLLMInstrumentor().instrument()
  1. Use module-level calls - Call litellm.completion(), not a directly imported completion() function

Enrich Your Traces

Add user IDs and custom metadata to traces

Custom Spans

Create spans for business logic around API calls

Distributed Tracing

Trace calls across service boundaries

​
Resources