OCT 2022
LangChain ships its first commit LangChain
Harrison Chase releases a thin wrapper around OpenAI with "chains." 0-to-10k GitHub stars in under 60 days — fastest open-source ML project to that milestone at the time.
DEC 2022
LlamaIndex (GPT Index) v0.1 LlamaIndex
Jerry Liu releases GPT Index, later renamed LlamaIndex. Purpose-built for retrieval and RAG — a deliberate architectural alternative to LangChain's general-purpose approach.
JAN 2023
LangChain raises $10M, dependency tree explodes LangChain
Rapid integration growth pushes transitive deps past 100. Speed of shipping beats architectural restraint. The 200+ dep footprint begins here.
JUN 2023
LlamaIndex 0.8 — enterprise feature push LlamaIndex
Advanced RAG features land: sub-question query engine, recursive retrieval, structured data support. Import time climbs as the core monolith grows without lazy loading.
JAN 2024
LangChain 0.1 — lazy loading architecture LangChain
Major refactor introduces deferred module loading. Import time drops from ~2.1s to under 0.6s. The dep count stays at 267 — lazy loading solves visibility, not surface area.
MAR 2024
SynapseKit 1.0 — 2-dependency design constraint SynapseKit
Built with a hard design rule: ship no transitive dependencies beyond httpx and pydantic. Trades ecosystem breadth for auditability, cold start speed, and zero version conflicts.
2024–2025
Framework fragmentation accelerates Ecosystem
Haystack, DSPy, CrewAI, AutoGen, Semantic Kernel all gain traction. The "pick a framework" decision gets harder as each optimizes for a different deployment model.