2
Home
Claude edited this page 2026-02-19 23:35:15 +00:00
Table of Contents
go-rag
Retrieval-Augmented Generation with vector search for the Lethean AI stack.
Module: forge.lthn.ai/core/go-rag
Size: ~1,017 LOC (7 files, 1 test file)
Coverage
go-rag: 18.4% (only chunk.go tested)
Phase Status
| Phase | Status | Notes |
|---|---|---|
| 0: Environment Setup | Partial | go.mod fixed; Qdrant + Ollama not running |
| 1: Unit Tests | Not started | Pure-function tests first, then service-dependent |
| 2: Test Infrastructure | Not started | Interface extraction for mocking |
| 3: Enhancements | Not started | Overlap, hybrid search, collection mgmt |
| 4: GPU Embeddings | Not started | ROCm Ollama, batch optimisation |
Critical Gaps
- Qdrant client: 226 lines, 0 tests — all gRPC calls, needs live Qdrant or mock
- Ollama client: 120 lines, 0 tests — Embed() needs live Ollama; EmbedDimension() is pure
- Ingest pipeline: 217 lines, 0 tests — orchestrates both services
- Query: 163 lines, 0 tests — FormatResults* functions are pure and testable now
Dependencies
| Module | Purpose |
|---|---|
core/go |
Logging (pkg/log) |
ollama/ollama |
Embedding API client |
qdrant/go-client |
Vector DB gRPC client |
Infrastructure Required
| Service | Status | Setup |
|---|---|---|
| Qdrant | Not running | docker run -d -p 6333:6333 -p 6334:6334 qdrant/qdrant |
| Ollama | Not running | curl -fsSL https://ollama.com/install.sh | sh && ollama pull nomic-embed-text |
Pages
- Chunking — Document splitting and category detection
- Vector-Search — Qdrant integration and query interface
- Test-Gaps — Detailed test suggestions per file
- Fleet-Context — How this repo fits in the agent fleet