qi-cli

name

qi-cli

description

Guide for using qi, a local knowledge search CLI for macOS and Linux. Use this skill whenever the user asks about qi commands, indexing documents, searching a knowledge base, asking questions with RAG, configuring providers (Ollama, OpenAI), understanding search modes (BM25, hybrid, vector), or anything related to the qi tool. Triggers on phrases like "qi index", "qi search", "qi ask", "qi query", "how do I use qi", "set up qi", "configure qi", "search my notes with qi", or any mention of the qi CLI.

qi is a local-first knowledge search CLI. It indexes documents into SQLite and supports BM25 full-text search, vector/hybrid search (with a local or remote embedding provider), and LLM-powered Q&A with citations.

Quick Start

qi init                                   # Create config + database
$EDITOR ~/.config/qi/config.yaml         # Add collections and (optionally) providers
qi index                                  # Index the current directory, or a collection
qi doctor                                 # Verify setup
qi search "your query"                    # BM25 keyword search (no provider needed)
qi query "your semantic question"         # Hybrid search (needs embedding provider)
qi ask "what does X do?"                  # RAG Q&A; use sparingly (needs generation provider)

---

Command selection guidance

Prefer qi index when the task is about adding, refreshing, or organizing source material. Prefer qi search or qi query when the task is about finding relevant documents, passages, or citations. Use qi ask sparingly, only when the user specifically needs a synthesized answer from an LLM rather than retrieved source results.

---

Commands

qi init

Writes ~/.config/qi/config.yaml (if absent) and initializes the SQLite database.

qi init

qi index [path|collection]

Indexes documents. SHA-256 content hashing means unchanged files are skipped. Embeddings are generated at index time — if you add an embedding provider to config after an initial index run, re-run qi index to populate the missing embeddings.

qi index                              # indexes current working directory
qi index ~/notes                      # any absolute or relative path
qi index notes                        # generated collection name from config

qi search <query>

BM25 full-text search. Fast, offline, no provider needed.

qi search "authentication"
qi search "deploy" -c code -n 5

qi query <query>

Semantic/hybrid search. Falls back gracefully to BM25 if the embedding provider is unavailable.

qi query "how does auth work"
qi query "deploy pipeline" --mode lexical   # BM25 only
qi query "deployment steps" --mode hybrid   # BM25 + vector (default)
qi query "critical path" --mode deep        # hybrid + reranking
qi query "question" --explain               # show BM25/vector/RRF score breakdown

Modes:

• lexical — BM25 only
• hybrid (default) — BM25 + vector fused with RRF; skips vector if BM25 has a clear winner
• deep — hybrid + optional reranking pass

qi ask <question>

RAG Q&A: searches the knowledge base, sends relevant chunks to an LLM, returns an answer with citations. Use this sparingly; prefer qi query for normal exploration, evidence gathering, and source lookup.

qi ask "What authentication methods are supported?"
qi ask "Explain the chunking algorithm" -c code

Requires a generation provider in config.

qi get <id>

Retrieve a document by its 6-character hash prefix (shown in search results).

qi get abc123

qi list

List all collections defined in config (name and path).

qi list

qi delete <collection>

Delete a collection: removes all indexed data from the database and removes the collection entry from config. Irreversible.

qi delete notes

qi stats

Show document counts, chunk counts, embedding counts, and database size per collection. If Embeddings count is much lower than Chunks count, vector search has no data to work with — re-run qi index.

qi doctor

Health-check config, database, collection paths, and provider connectivity. If the embedding provider shows SKIP instead of OK, qi query will silently fall back to BM25 regardless of your config.

qi update

Update the binary from GitHub. If installed via Homebrew, it suggests brew upgrade qi instead.

---

Global Flags

Flag	Description
-v, --verbose	Verbose/debug output
-f, --format text|json|markdown	Output format (default: text)
--config <path>	Override config path
-c, --collection <name>	Limit to a specific collection
-n, --limit <N>	Number of results (default: 10)

---

Configuration (~/.config/qi/config.yaml)

database_path: ~/.local/share/qi/qi.db

collections:
  - name: notes
    path: ~/notes
    extensions: [.md, .txt]
    ignore: [.git]

  - name: code
    path: ~/Projects/myproject
    extensions: [.go, .ts, .py]
    ignore: [vendor, dist]

providers:
  embedding:                              # optional — enables vector/hybrid search
    base_url: http://localhost:11434      # Ollama, LM Studio, llama.cpp, OpenAI-compatible
    model: nomic-embed-text
    dimension: 768

  rerank:                                 # optional — enables deep mode
    base_url: http://localhost:8080
    model: bge-reranker-v2-m3

  generation:                             # optional — enables `qi ask`
    base_url: http://localhost:11434
    model: llama3.2
    api_key: ""                           # set for OpenAI or auth-gated services

search:
  default_mode: hybrid                    # lexical | hybrid | deep
  bm25_top_k: 50
  vector_top_k: 50
  rerank_top_k: 10
  rrf_k: 60
  chunk_size: 512

Common provider setups

Ollama (fully local):

providers:
  embedding:
    base_url: http://localhost:11434
    model: nomic-embed-text
    dimension: 768
  generation:
    base_url: http://localhost:11434
    model: llama3.2

OpenAI:

providers:
  embedding:
    base_url: https://api.openai.com/v1
    model: text-embedding-3-small
    dimension: 1536
  generation:
    base_url: https://api.openai.com/v1
    model: gpt-4o
    api_key: sk-...

Mixed (local embeddings, cloud generation):

providers:
  embedding:
    base_url: http://localhost:11434
    model: nomic-embed-text
    dimension: 768
  generation:
    base_url: https://api.openai.com/v1
    model: gpt-4o
    api_key: sk-...

---

How search works

• BM25 — SQLite FTS5. Always available, very fast, good for keyword queries.
• Vector KNN — Cosine similarity over embedding BLOBs in SQLite. Requires an embedding provider. Captures semantic intent.
• Hybrid (RRF) — Runs both, fuses rankings with Reciprocal Rank Fusion (score = Σ 1/(k + rank)). Skips vector if BM25 has a dominant winner (top score > 3× second place).
• Deep — Hybrid + a second-pass reranker for best accuracy.

---

Document references

Search results show locations like qi://notes/2024/jan.md [Section > Subsection] and a 6-character ID. Use qi get <id> to view the full document.

---

Typical workflows

Index and search (no provider needed):

qi init
qi index ~/notes                 # saves and indexes ~/notes as "notes"
qi search "my keyword" -c notes

Manage collections:

qi list                          # see all configured collections
qi index ~/projects              # add a new collection
qi delete old-notes              # remove collection data + config entry

Semantic search with Ollama:

# pull a model: ollama pull nomic-embed-text
# add embedding provider to config
qi index notes
qi query "how does X work" --explain

RAG Q&A (use sparingly):

# also add a generation provider to config
# prefer qi query unless you need a synthesized answer
qi ask "Summarize the key decisions in my notes"

Debug / inspect:

qi doctor                         # check all providers are reachable
qi stats                          # see document/chunk/embedding counts
qi query "question" --explain     # see score breakdown
qi get abc123                     # read the full source document

---

Troubleshooting

qi query returns keyword-only results despite embedding provider being configured

Work through these steps in order:

1. Check if the provider is actually wired up

qi doctor

Look for OK next to the embedding provider. SKIP means qi didn't parse the provider config — common causes: wrong YAML key name (embedding not embeddings), missing dimension field, or bad indentation (must be under providers:).

2. Check if embeddings were actually generated

qi stats

If Embeddings is 0 or far less than Chunks, the vectors were never written. This happens when the embedding provider was added to config after the initial qi index run — re-index to fix it:

qi index <collection-name>

Only chunks missing embeddings are re-processed; unchanged files aren't re-parsed.

3. Verify the provider is reachable at query time qi query embeds the query string at runtime. If the provider is down, qi silently falls back to BM25 without an error. Test connectivity:

# Ollama:
curl http://localhost:11434/v1/embeddings -H "Content-Type: application/json" \
  -d '{"model":"nomic-embed-text","input":["test"]}'

Also confirm the model is pulled: ollama list.

4. Use --explain to see what's actually happening

qi query "your question" --explain

This shows each result's BM25 rank, vector rank, and fused RRF score. If every result is missing a vector rank, the vector path isn't contributing — confirming one of the issues above.

Note: hybrid mode also skips vector search if the top BM25 score is more than 3× the second result (intentional optimization for dominant keyword matches). If your queries are exact keyword matches, try a more paraphrased, natural-language query to avoid this shortcut.