Research pass flagged retrieval mismatch as a recurring vector/RAG failure: similar is not always relevant, true or authorized.
Tool profile · provisional profile
Chroma
Useful embedding store. It is infrastructure for recall, not a complete memory architecture.
Provisional fit
58/100
Best for: Builders who need a straightforward vector substrate and will build source, curation and governance around it.
Avoid if: you need a fully governed, citation-complete knowledge architecture without adding policy, evidence capture, and review workflow around the tool.
Caution: Similarity search can accidentally become policy. Authority, freshness, privacy and deletion must live above the vector DB.
Model signature: Storage primary · app scope · Vector DB
Layer coverage
Where this tool fits.
This is not a completed review. It is a provisional profile from public positioning plus known failure-mode mapping. Hands-on benchmarks, source snapshots, and citation-bound claims are still required before stronger conclusions.
Production
Curation
Storage
Context Assembly
Governance
Review rule: a tool does not get credit for a layer unless it exposes inspectable behavior, not just a marketing claim.
Evidence notes
What the provisional profile has applied so far.
Mapped to storage because Chroma provides substrate, not end-to-end knowledge lifecycle.
Needs review for metadata filtering, delete/update behavior, multi-tenant isolation and observability.
Review packet
What a complete review must contain.
This page exposes the intended review structure. The current artifact is a profile, not a completed evidence-backed review.
Canonical source
Strengths
Builders who need a straightforward vector substrate and will build source, curation and governance around it.
Limitations
Similarity search can accidentally become policy. Authority, freshness, privacy and deletion must live above the vector DB.
Dimension assessment
Scope, volatility, authority, lifecycle, resource economics, interoperability, and evidence quality must each get a rationale and citations before final scoring.
Open questions
- What can be verified from docs, code, issues, benchmarks, and changelogs?
- Where does the tool fail under stale, contradictory, private, or high-cost knowledge?
- Which claims are vendor claims versus independently observed behavior?
Benchmark critique
No benchmark number is accepted as architectural evidence unless it says which layer it tests and what it misses: lifecycle, scope boundaries, authority, context cost, and governance.
Related systems
Related tools should be connected by evidence-backed edges: competes with, integrates with, implements concept, evaluated by, or has governance gap.
Update history
Provisional profile created. Stale-review detection, source snapshots, and changelog watching are required before this becomes a durable review.