langtrace

// Automatic tracing — one line setup
import * as Langtrace from "@langtrase/typescript-sdk";

Langtrace.init({
  api_key: process.env.LANGTRACE_API_KEY,
  batch: true,                            // Batch spans for performance
  instrumentations: {
    openai: true,
    anthropic: true,
    langchain: true,
    pinecone: true,
    chromadb: true,
  },
});

// Now all LLM calls are automatically traced
import OpenAI from "openai";
const openai = new OpenAI();

// This call is automatically instrumented — no code changes needed
const response = await openai.chat.completions.create({
  model: "gpt-4o",
  messages: [{ role: "user", content: "Explain quantum computing" }],
});
// Langtrace captures: model, tokens, latency, cost, input, output

from langtrace_python_sdk import langtrace, with_langtrace_root_span
from langtrace_python_sdk.utils.with_root_span import with_additional_attributes

langtrace.init(api_key="your-api-key")

@with_langtrace_root_span("rag-pipeline")
async def answer_question(query: str) -> str:
    """Full RAG pipeline — every step traced automatically."""

    # Step 1: Embed query (traced)
    embedding = await openai.embeddings.create(
        model="text-embedding-3-small", input=query,
    )

    # Step 2: Vector search (traced if Pinecone/Chroma instrumented)
    results = index.query(vector=embedding.data[0].embedding, top_k=5)

    # Step 3: Generate answer (traced)
    context = "\n".join([r.metadata["text"] for r in results.matches])
    response = await openai.chat.completions.create(
        model="gpt-4o",
        messages=[
            {"role": "system", "content": f"Answer using this context:\n{context}"},
            {"role": "user", "content": query},
        ],
    )

    return response.choices[0].message.content

# Custom attributes for filtering
@with_additional_attributes({"user.tier": "pro", "feature": "search"})
@with_langtrace_root_span("search")
async def pro_search(query: str):
    return await answer_question(query)

from langtrace_python_sdk import langtrace
from langtrace_python_sdk.utils.with_root_span import with_langtrace_root_span

# Run evaluations and track scores
@with_langtrace_root_span("evaluate-rag")
async def evaluate_rag(test_set: list[dict]):
    results = []
    for test in test_set:
        answer = await answer_question(test["query"])

        # Score with LLM-as-judge
        eval_response = await openai.chat.completions.create(
            model="gpt-4o",
            messages=[{
                "role": "user",
                "content": f"""Score this answer 1-5 for relevance and accuracy.
                Question: {test['query']}
                Expected: {test['expected']}
                Got: {answer}
                Return JSON: {{"relevance": N, "accuracy": N}}""",
            }],
        )
        scores = json.loads(eval_response.choices[0].message.content)
        results.append({**scores, "query": test["query"]})

    # All evaluation traces visible in Langtrace dashboard
    avg_relevance = sum(r["relevance"] for r in results) / len(results)
    avg_accuracy = sum(r["accuracy"] for r in results) / len(results)
    return {"avg_relevance": avg_relevance, "avg_accuracy": avg_accuracy}

# TypeScript
npm install @langtrase/typescript-sdk

# Python
pip install langtrace-python-sdk

# Self-hosted
docker run -p 3000:3000 langtrace/langtrace-client