feat: Modular dependency installation for langflow-base

[ogabrielluiz]

Summary

Makes langflow-base dependencies modular, allowing users to install only what they need instead of the full dependency tree.

For users who want everything (current behavior):

pip install langflow  # installs langflow-base[complete] automatically

For users who want a minimal installation:

pip install langflow-base[openai,chroma]  # only OpenAI and Chroma
pip install langflow-base[google,pinecone,langfuse]  # custom stack

Available Optional Groups

Category	Options
Databases	couchbase, cassandra, clickhouse, mongodb, supabase, redis, elasticsearch, postgresql
Vector Stores	faiss, qdrant, weaviate, chroma, upstash, pinecone, milvus, mongodb-vectors, astradb
LLM Providers	openai, anthropic, cohere, google, ollama, nvidia, mistral, sambanova, groq
Local LLMs	llama-cpp, sentence-transformers, ctransformers
Monitoring	langfuse, langwatch, langsmith, arize
Agent Frameworks	dspy, smolagents
Tools	beautifulsoup, serpapi, google-api, wikipedia, duckduckgo, yfinance, wolframalpha
Complete	complete or all (installs everything)

Why?

• Faster installation for users who don't need all integrations
• Smaller Docker images for production deployments
• Fewer dependency conflicts when only specific integrations are needed
• Clearer understanding of what each integration requires

[codspeed-hq]

CodSpeed Performance Report

Merging #6732 will improve performances by 65.1%

_{Comparing better-langflow-base (fd380f4) with main (0886bda)}

Summary

⚡ 1 improvements
✅ 13 untouched benchmarks

Benchmarks breakdown

	Benchmark	BASE	HEAD	Change
⚡	test_build_flow_invalid_job_id	15.2 ms	9.2 ms	+65.1%