airweave

Files (5)

• SKILL.md
• _meta.json
• EXAMPLES.md · references/
• PARAMETERS.md · references/
• search.py · scripts/

SKILL.md (first lines)

# Airweave Search

Search and retrieve context from Airweave collections using the search script at `{baseDir}/scripts/search.py`.

## When to Search

**Search when the user:**
- Asks about data in their connected apps ("What did we discuss in Slack about...")
- Needs to find documents, messages, issues, or records
- Asks factual questions about their workspace ("Who is responsible for...", "What's our policy on...")
- References specific tools by name ("in Notion", "on GitHub", "in Jira")
- Needs recent information you don't have in your training
- Needs you to check app data for context ("check our Notion docs", "look at the Jira ticket")

**Don't search when:**
- User asks general knowledge questions (use your training)
- User already provided all needed context in the conversation
- The question is about Airweave itself, not data within it

## Query Formulation

Turn user intent into effective search queries:

| User Says | Search Query |
|-----------|--------------|
| "What did Sarah say about the launch?" | "Sarah product launch" |
| "Find the API documentation" | "API documentation" |
| "Any bugs reported this week?" | "bug report issues" |
| "What's our refund policy?" | "refund policy customer" |

**Tips:**
- Use natural language — Airweave uses semantic search
- Include context — "pricing feedback" beats just "pricing"
- Be specific but not too narrow
- Skip filler words like "please find", "can you search for"

## Running a Search

Execute the search script:

```bash
python3 {baseDir}/scripts/search.py "your search query"
```

**Optional parameters:**
- `--limit N` — Max results (default: 20)
- `--temporal N` — Temporal relevance 0-1 (default: 0, use 0.7+ for "recent", "latest")
- `--strategy TYPE` — Retrieval strategy: hybrid, semantic, keyword (default: hybrid)
- `--raw` — Return raw results instead of AI-generated answer
- `--expand` — Enable query expansion for broader results
- `--rerank / --no-rerank` — Toggle LLM reranking (default: on)

**Examples:**

```bash
# Basic search
python3 {baseDir}/scripts/search.py "customer feedback pricing"

# Recent conversations
python3 {baseDir}/scripts/search.py "product launch updates" --temporal 0.8

# Find specific document
python3 {baseDir}/scripts/search.py "API authentication docs" --strategy keyword

# Get raw results for exploration
python3 {baseDir}/scripts/search.py "project status" --limit 30 --raw

# Broad search with query expansion
python3 {baseDir}/scripts/search.py "onboarding" --expand
```

## Handling Results

**Interpreting scores:**
- 0.85+ → Highly relevant, use confidently
- 0.70-0.85 → Likely relevant, use with context
- 0.50-0.70 → Possibly relevant, mention uncertainty
- Below 0.50 → Weak match, consider rephrasing

**Presenting to users:**
1. Lead with the answer — don't start with "I found 5 results"
2. Cite sources — mention where info came from ("According to your Slack conversation...")
3. Synthesize — combine relevant parts into a coherent response
4. Acknowledge gaps — if results don't fully answer, say so

## Handling No Results

If search returns nothing useful:
1. Broaden the query — remove specific terms
2. Try different phrasing — use synonyms
3. Increase limit — fetch more results
4. Ask for clarification — user might have more context

## Parameter Reference

See [PARAMETERS.md](references/PARAMETERS.md) for detailed parameter guidance.

## Examples

See [EXAMPLES.md](references/EXAMPLES.md) for complete search scenarios.

From CraftOS-dev/CraftBot, MIT.

More from CraftOS-dev/CraftBot