langchain-brime 0.2.0

LangChain integration for the Brime API (search · extract · research) — tools and retriever

langchain-brime

Brime API · Docs · LangChain guide · Python SDK · Issues

LangChain meets the live web. Drop the Brime API into any LangChain agent, chain, or RAG pipeline — as three typed tools and a retriever you can plug in unchanged.

pip install langchain-brime

Python 3.9+. Sync and async. Fully typed (py.typed). SecretStr API keys. Actionable agent errors via handle_tool_error. Passes the official LangChain ToolsUnitTests compliance suite.

What you get

Class	Type	What it does
BrimeSearch	BaseTool	Fast SERP search with snippets + synthesised answer.
BrimeExtract	BaseTool	Any URL → clean markdown (HTML / PDF / DOCX / SPA).
BrimeResearch	BaseTool	Multi-round agent research with citations.
BrimeRetriever	BaseRetriever	Live-web retriever for RAG — returns LangChain Documents.

One API key powers all four. One set of tuned defaults. No knobs to babysit.

30 seconds

from langchain.agents import create_agent
from langchain_brime import BrimeSearch, BrimeExtract

agent = create_agent(
    model="anthropic:claude-3-5-haiku-latest",
    tools=[BrimeSearch(), BrimeExtract()],
    system_prompt="You are a research assistant.",
)

result = agent.invoke({
    "messages": [{"role": "user", "content": "What is BM25 ranking? Cite sources."}],
})
print(result["messages"][-1].content)

RAG without the RAG plumbing

BrimeRetriever returns standard LangChain Documents — drop it into any retrieval chain and you're querying the live web with BM25-ranked chunks:

from langchain_brime import BrimeRetriever
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.runnables import RunnablePassthrough
from langchain.chat_models import init_chat_model

retriever = BrimeRetriever(k=5)
llm = init_chat_model("anthropic:claude-3-5-haiku-latest")

prompt = ChatPromptTemplate.from_template(
    "Context:\n{context}\n\nQuestion: {question}\n\nAnswer with citations."
)
chain = ({"context": retriever, "question": RunnablePassthrough()} | prompt | llm)

print(chain.invoke("How does BM25 weight term frequency?").content)

Each Document carries page_content (ranked chunk) and rich metadata: url, title, score, published_date, source="brime".

Async, end-to-end

import asyncio
from langchain_brime import BrimeSearch

async def main():
    result = await BrimeSearch().ainvoke({"query": "python async io patterns"})
    print(result)

asyncio.run(main())

Every tool has _arun. The retriever has _aget_relevant_documents. No surprises.

Authentication

export BRIME_API_KEY="sk-brime-..."

BrimeSearch()                        # uses BRIME_API_KEY
BrimeSearch(api_key="sk-brime-...")  # explicit
BrimeSearch(base_url="https://...")  # staging override (or BRIME_BASE_URL)

Get a key at brime.dev — free tier: 1,000 credits/month, no card.

Tool reference

BrimeSearch

Field	Default	Notes
query	(required)	Natural-language query
topic	"general"	general · news · finance (recency hint)
max_results	5	1–20
time_range	None	day · week · month · year

Returns a markdown string with a synthesised answer plus a Sources: list.

BrimeExtract

Field	Default	Notes
urls	(req.)	List of URLs (1–10)
include_metadata	False	Add extra metadata fields per result

Returns markdown with a ## <url> heading per page, plus a ## Failed URLs block if any failed.

BrimeResearch

Field	Default	Notes
query	(required)	Research question
topic	"general"	Recency hint
max_results	5	Sources per round, 1–20

Single tuned multi-round flow (~30–90 s) with cited synthesis.

BrimeRetriever

Field	Default	Notes
k	5	Max documents (1–20)
topic	"general"	Recency hint
time_range	None	Recency window
domains	None	Domain allowlist
exclude_domains	None	Domain denylist
include_answer	False	Skip the LLM answer when used purely for retrieval

Why this instead of writing your own wrapper?

• One package, four classes. Search, extract, research, and retriever — all sharing one HTTP client (brime SDK).
• Idempotency, handled. Auto Idempotency-Key for extract (Brime API requirement, easy to miss).
• Async parity throughout. No partially-async surface area.
• LangChain-tested. Passes the official ToolsUnitTests compliance suite (args schema serialisable, async parity, no override violations).

Compatibility

• Python 3.9+
• langchain-core ≥ 1.4, < 2.0
• Backed by brime Python SDK ≥ 0.2.0 (retry, request_id, typed error hierarchy)

Links

• Docs — docs.brime.dev
• LangChain integration guide — docs.brime.dev/sdks/langchain
• Status — brime.dev
• Issues — github.com/brime-dev/brime-langchain/issues

License

MIT © Brime