bmlib

Shared Python library for biomedical literature tools — LLM abstraction, quality assessment, transparency analysis, full-text retrieval, publication ingestion, and database utilities.

Version: 0.6.0 | License: AGPL-3.0-or-later | Python: >=3.11

Installation

# Core (only jinja2 dependency)
pip install bmlib

# Editable install with all extras
uv pip install -e ".[all,dev]"

Optional dependency groups

Modules

Quick Start

Database

from bmlib.db import connect_sqlite, execute, fetch_all, transaction

conn = connect_sqlite("~/.myapp/data.db")
with transaction(conn):
    execute(conn, "INSERT INTO papers (doi, title) VALUES (?, ?)", ("10.1101/x", "A paper"))
rows = fetch_all(conn, "SELECT * FROM papers")

LLM

from bmlib.llm import LLMClient, LLMMessage

client = LLMClient(default_provider="ollama")
response = client.chat(
    messages=[LLMMessage(role="user", content="Summarise this paper.")],
    model="ollama:medgemma4B_it_q8",
)
print(response.content)

Model strings use the format "provider:model_name":

"anthropic:claude-sonnet-4-20250514"
"openai:gpt-4o"
"ollama:medgemma4B_it_q8"
"deepseek:deepseek-chat"
"mistral:mistral-large-latest"
"gemini:gemini-2.0-flash"

Tool Calling

from bmlib.llm import LLMClient, LLMMessage, LLMToolDefinition

search = LLMToolDefinition(
    name="search_pubmed",
    description="Search PubMed for articles matching a query.",
    parameters={
        "type": "object",
        "properties": {"query": {"type": "string"}},
        "required": ["query"],
    },
)

client = LLMClient()
response = client.chat(
    messages=[LLMMessage(role="user", content="Find recent trials on statins.")],
    model="anthropic:claude-sonnet-4-20250514",
    tools=[search],
)

for call in response.tool_calls or []:
    print(call.name, call.arguments)  # arguments is already a parsed dict

To continue the conversation, append the assistant message (carrying tool_calls) and one role="tool" message per call, each with the matching tool_call_id, then send the whole list again.

Long Documents

from bmlib.llm import chunk_text, process_with_map_reduce

for chunk in chunk_text(paper_text, chunk_size=8000, overlap=200):
    print(chunk.chunk_index, chunk.size)

summary = process_with_map_reduce(
    paper_text,
    map_fn=lambda part: summarise(part),
    reduce_fn=lambda parts: summarise("\n".join(parts)),
)

Publication Sync

from datetime import date
from bmlib.db import connect_sqlite
from bmlib.publications import sync

conn = connect_sqlite("publications.db")
report = sync(
    conn,
    sources=["pubmed", "biorxiv"],
    date_from=date(2025, 1, 1),
    date_to=date(2025, 1, 7),
    email="researcher@example.com",
)
print(f"Added: {report.records_added}, Merged: {report.records_merged}")

Full-Text Retrieval

from bmlib.fulltext import FullTextService

service = FullTextService(email="researcher@example.com")

# Passing identifier= enables the built-in disk cache (platform default dir).
result = service.fetch_fulltext(
    pmc_id="PMC7614751", doi="10.1234/example", identifier="PMC7614751"
)

if result.html:
    print(result.html[:200])

Quality Assessment

from bmlib.llm import LLMClient
from bmlib.quality import QualityManager

llm = LLMClient()
manager = QualityManager(
    llm=llm,
    classifier_model="anthropic:claude-3-haiku-20240307",
    assessor_model="anthropic:claude-sonnet-4-20250514",
)

assessment = manager.assess(
    title="A Randomized Controlled Trial of ...",
    abstract="We conducted a double-blind RCT ...",
    publication_types=["Randomized Controlled Trial"],
)
print(assessment.study_design, assessment.quality_tier)

Transparency Analysis

from bmlib.transparency import TransparencyAnalyzer

analyzer = TransparencyAnalyzer(email="researcher@example.com")
result = analyzer.analyze("doc-001", doi="10.1038/s41586-024-00001-0")
print(result.transparency_score, result.risk_level)

Development

# Install with dev dependencies
uv pip install -e ".[all,dev]"

# Run tests
uv run pytest tests/ -v

# Lint and format
uv run ruff check .
uv run ruff format --check .

Documentation

Full API documentation is available in docs/manual/.

License

AGPL-3.0-or-later