v0.0.1

docsiq v0.0.1 — first stable release

First non-beta release of docsiq after an extended beta phase. This
release establishes the feature set and API surface that subsequent
0.0.x patches will maintain back-compat against.

What's in the box

• GraphRAG indexing pipeline — five-phase ingestion: chunk → extract
  entities + relationships + claims → community-detect (Louvain) →
  embed → persist.
• Document loaders — PDF (langchaingo), DOCX, TXT, Markdown, and a
  polite web crawler with robots.txt + allow-list + MIME checks.
• Multi-provider LLM layer — Azure OpenAI, OpenAI, and Ollama behind
  a single internal/llm abstraction (langchaingo underneath).
• Query engine — hybrid local (vector + FTS5) and global
  (community-summary) search.
• Surfaces — CLI (docsiq index|search|serve), REST API, MCP
  server, and an embedded React SPA served by docsiq serve.
• Storage — single SQLite file with sqlite_fts5 + sqlite-vec
  for vector search. No external DB to deploy.

Install

# Signed binary
gh release download v0.0.1 -p 'docsiq-v0.0.1-linux-amd64*'

# Or from source
go install github.com/RandomCodeSpace/docsiq@v0.0.1

Upgrade impact

No previous stable release exists — this is v0.0.1. Users upgrading
from v0.0.0-beta.* should start with a fresh data directory; the
schema is the same as the final beta but the beta tags have been
retired.

Security

• All release binaries are signed with cosign keyless via Sigstore
  and anchored to the Rekor transparency log.
• A signed SHA256SUMS is included. Verification instructions are
  attached to this release.
• SLSA build provenance (.intoto.jsonl) accompanies the binaries.

Known limitations

• Darwin support is limited to arm64; amd64 binaries are not
  built (cgo + sqlite-vec cross-compile complexity).
• Pre-1.0: APIs and on-disk schema are not yet frozen.