Your GitHub commit history is your best marketing copy — unwritten
LangChain pipelines that generate developer brand copy from GitHub activity.
Most developers can't articulate what they actually build in terms non-engineers understand. BrandForce runs LangChain pipelines over your commit history, README files, and issue patterns to generate positioning copy, bio text, and technical writeups that reflect your real work — not a recycled job description.
The challenge
Transforming sparse, inconsistent commit messages and README prose into coherent, audience-calibrated brand copy without hallucinating capabilities the developer doesn't have.
The outcome
Developers get a positioning brief, short bio, and project writeups generated directly from their public GitHub — ready to paste into portfolios, LinkedIn, and proposal decks.
Technical approach
- LangChain pipeline — composable chain of LLM calls with structured output parsing
- GitHub API ingestion — commits, README, and issue data fetched and normalised per repo
- Supabase Edge Functions — Deno-based serverless functions for pipeline execution
- Vector embeddings — commit messages and README chunks embedded for semantic clustering
- Supabase pgvector — PostgreSQL extension for embedding storage and similarity search
- Structured output parsing — LangChain output parsers enforce typed brand copy schema
- Prompt chaining — multi-step prompts: summarise → position → tone-match → finalise
- OpenAI function calling — constrained generation for consistent copy format output
Results at a glance
- In development
- Status
- LangChain
- Pipeline