[APP] Build a Chroma-powered semantic grep utility (local + cloud)

[tazarov]

Why

A grep-like app is a compelling way to demonstrate Chroma’s practical value beyond basic demos. Users understand grep immediately, so a semantic/hybrid “next-gen grep” makes adoption easier.

Proposal

Build a chroma-grep demo app that supports searching for information locally (files/repos) and remotely (cloud-backed indexes/services).

Core UX

• CLI-first workflow with familiar grep ergonomics.
• Query modes:
  • exact (keyword/regex)
  • semantic (vector similarity)
  • hybrid (keyword + semantic fusion)
• Output includes match snippet, source path/service, score, and metadata filters.

Scope (MVP)

• Local indexing/search for text files and markdown.
• Chroma-backed semantic index + optional lexical baseline.
• Cloud mode via configurable endpoint (HTTP API) for remote corpora.
• --mode flag (exact|semantic|hybrid) and --source flag (local|cloud|both).
• Include examples for “find API auth failures”, “find PII handling mentions”, etc.

Stretch Scope

• Incremental indexing and file-watch updates.
• Query-time reranking.
• Team-shared cloud memory spaces.

Deliverables

• Runnable CLI app in cookbook examples.
• End-to-end tutorial: index -> query -> compare modes.
• Benchmark/eval section comparing exact vs semantic vs hybrid quality.
• Docker compose profile for cloud-backed demo.

Acceptance Criteria

• User can install and run with documented commands.
• Same query can run against local-only, cloud-only, or combined sources.
• Hybrid mode shows measurable improvement on at least one provided eval set.
• Tutorial includes troubleshooting and production extension notes.