Multi-source retrieval with confidence scoring - web, academic, and Tavily in one unified API

One API key. Multi-source retrieval. Confidence-scored answers.

> Inspired by [AIsa Verity](https://github.com/AIsa-team/verity) - A next-generation search agent with trust-scored answers.

## 🔥 What Can You Do?

### Research Assistant
```
"Search for the latest papers on transformer architectures from 2024-2025"
```

### Market Research
```
"Find all web articles about AI startup funding in Q4 2025"
```

### Competitive Analysis
```
"Search for reviews and comparisons of RAG frameworks"
```

### News Aggregation
```
"Get the latest news about quantum computing breakthroughs"
```

### Deep Dive Research
```
"Smart search combining web and academic sources on 'autonomous agents'"
```

## Quick Start

```bash
export AISA_API_KEY="your-key"
```

---

## 🏗️ Architecture: Multi-Stage Orchestration

OpenClaw Search employs a **Two-Phase Retrieval Strategy** for comprehensive results:

### Phase 1: Discovery (Parallel Retrieval)

Query 4 distinct search streams simultaneously:
- **Scholar**: Deep academic retrieval
- **Web**: Structured web search
- **Smart**: Intelligent mixed-mode search
- **Tavily**: External validation signal

### Phase 2: Reasoning (Meta-Analysis)

Use **AIsa Explain** to perform meta-analysis on search results, generating:
- Confidence scores (0-100)
- Source agreement analysis
- Synthesized answers

```
┌─────────────────────────────────────────────────────────────┐
│                      User Query                              │
└─────────────────────────────────────────────────────────────┘
                              │
              ┌───────────────┼───────────────┐
              ▼               ▼               ▼
        ┌─────────┐     ┌─────────┐     ┌─────────┐
        │ Scholar │     │   Web   │     │  Smart  │
        └─────────┘     └─────────┘     └─────────┘
              │               │               │
              └───────────────┼───────────────┘
                              ▼
                    ┌─────────────────┐
                    │  AIsa Explain   │
                    │ (Meta-Analysis) │
                    └─────────────────┘
                              │
                              ▼
                    ┌─────────────────┐
                    │ Confidence Score│
                    │  + Synthesis    │
                    └─────────────────┘
```

---

## Core Capabilities

### Web Search

```bash
# Basic web search
curl -X POST "https://api.aisa.one/apis/v1/scholar/search/web?query=AI+frameworks&max_num_results=10" \
  -H "Authorization: Bearer $AISA_API_KEY"

# Full text search (with page content)
curl -X POST "https://api.aisa.one/apis/v1/search/full?query=latest+AI+news&max_num_results=10" \
  -H "Authorization: Bearer $AISA_API_KEY"
```

### Academic/Scholar Search

```bash
# Search academic papers
curl -X POST "https://api.aisa.one/apis/v1/scholar/search/scholar?query=transformer+models&max_num_results=10" \
  -H "Authorization: Bearer $AISA_API_KEY"

# With year filter
curl -X POST "https://api.aisa.one/apis/v1/scholar/search/scholar?query=LLM&max_num_results=10&as_ylo=2024&as_yhi=2025" \
  -H "Authorization: Bearer $AISA_API_KEY"
```

### Smart Search (Web + Academic Combined)

```bash
# Intelligent hybrid search
curl -X POST "https://api.aisa.one/apis/v1/scholar/search/smart?query=machine+learning+optimization&max_num_results=10" \
  -H "Authorization: Bearer $AISA_API_KEY"
```

### Tavily Integration (Advanced)

```bash
# Tavily search
curl -X POST "https://api.aisa.one/apis/v1/tavily/search" \
  -H "Authorization: Bearer $AISA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"query":"latest AI developments"}'

# Extract content from URLs
curl -X POST "https://api.aisa.one/apis/v1/tavily/extract" \
  -H "Authorization: Bearer $AISA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"urls":["https://example.com/article"]}'

# Crawl web pages
curl -X POST "https://api.aisa.one/apis/v1/tavily/crawl" \
  -H "Authorization: Bearer $AISA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"url":"https://example.com","max_depth":2}'

# Site map
curl -X POST "https://api.aisa.one/apis/v1/tavily/map" \
  -H "Authorization: Bearer $AISA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"url":"https://example.com"}'
```

### Explain Search Results (Meta-Analysis)

```bash
# Generate explanations with confidence scoring
curl -X POST "https://api.aisa.one/apis/v1/scholar/explain" \
  -H "Authorization: Bearer $AISA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"results":[...],"language":"en","format":"summary"}'
```

---

## 📊 Confidence Scoring Engine

Unlike standard RAG systems, OpenClaw Search evaluates credibility and consensus:

### Scoring Rubric

| Factor | Weight | Description |
|--------|--------|-------------|
| **Source Quality** | 40% | Academic > Smart/Web > External |
| **Agreement Analysis** | 35% | Cross-source consensus checking |
| **Recency** | 15% | Newer sources weighted higher |
| **Relevance** | 10% | Query-result semantic match |

### Score Interpretation

| Score | Confidence Level | Meaning |
|-------|-----------------|---------|
| 90-100 | Very High | Strong consensus across academic and web sources |
| 70-89 | High | Good agreement, reliable sources |
| 50-69 | Medium | Mixed signals, verify independently |
| 30-49 | Low | Conflicting sources, use caution |
| 0-29 | Very Low | Insufficient or contradictory data |

---

## Python Client

```bash
# Web search
python3 {baseDir}/scripts/search_client.py web --query "latest AI news" --count 10

# Academic search
python3 {baseDir}/scripts/search_client.py scholar --query "transformer architecture" --count 10
python3 {baseDir}/scripts/search_client.py scholar --query "LLM" --year-from 2024 --year-to 2025

# Smart search (web + academic)
python3 {baseDir}/scripts/search_client.py smart --query "autonomous agents" --count 10

# Full text search
python3 {baseDir}/scripts/search_client.py full --query "AI startup funding"

# Tavily operations
python3 {baseDir}/scripts/search_client.py tavily-search --query "AI developments"
python3 {baseDir}/scripts/search_client.py tavily-extract --urls "https://example.com/article"

# Multi-source search with confidence scoring
python3 {baseDir}/scripts/search_client.py verity --query "Is quantum computing ready for enterprise?"
```

---

## API Endpoints Reference

| Endpoint | Method | Description |
|----------|--------|-------------|
| `/scholar/search/web` | POST | Web search with structured results |
| `/scholar/search/scholar` | POST | Academic paper search |
| `/scholar/search/smart` | POST | Intelligent hybrid search |
| `/scholar/explain` | POST | Generate result explanations |
| `/search/full` | POST | Full text search with content |
| `/search/smart` | POST | Smart web search |
| `/tavily/search` | POST | Tavily search integration |
| `/tavily/extract` | POST | Extract content from URLs |
| `/tavily/crawl` | POST | Crawl web pages |
| `/tavily/map` | POST | Generate site maps |

---

## Search Parameters

| Parameter | Type | Description |
|-----------|------|-------------|
| query | string | Search query (required) |
| max_num_results | integer | Max results (1-100, default 10) |
| as_ylo | integer | Year lower bound (scholar only) |
| as_yhi | integer | Year upper bound (scholar only) |

---

## 🚀 Building a Verity-Style Agent

Want to build your own confidence-scored search agent? Here's the pattern:

### 1. Parallel Discovery

```python
import asyncio

async def discover(query):
    """Phase 1: Parallel retrieval from multiple sources."""
    tasks = [
        search_scholar(query),
        search_web(query),
        search_smart(query),
        search_tavily(query)
    ]
    results = await asyncio.gather(*tasks)
    return {
        "scholar": results[0],
        "web": results[1],
        "smart": results[2],
        "tavily": results[3]
    }
```

### 2. Confidence Scoring

```python
def score_confidence(results):
    """Calculate deterministic confidence score."""
    score = 0
    
    # Source quality (40%)
    if results["scholar"]:
        score += 40 * len(results["scholar"]) / 10
    
    # Agreement analysis (35%)
    claims = extract_claims(results)
    agreement = analyze_agreement(claims)
    score += 35 * agreement
    
    # Recency (15%)
    recency = calculate_recency(results)
    score += 15 * recency
    
    # Relevance (10%)
    relevance = calculate_relevance(results, query)
    score += 10 * relevance
    
    return min(100, score)
```

### 3. Synthesis

```python
async def synthesize(query, results, score):
    """Generate final answer with citations."""
    explanation = await explain_results(results)
    return {
        "answer": explanation["summary"],
        "confidence": score,
        "sources": explanation["citations"],
        "claims": explanation["claims"]
    }
```

For a complete implementation, see [AIsa Verity](https://github.com/AIsa-team/verity).

---

## Pricing

| API | Cost |
|-----|------|
| Web search | ~$0.001 |
| Scholar search | ~$0.002 |
| Smart search | ~$0.002 |
| Tavily search | ~$0.002 |
| Explain | ~$0.003 |

Every response includes `usage.cost` and `usage.credits_remaining`.

---

## Get Started

1. Sign up at [aisa.one](https://aisa.one)
2. Get your API key
3. Add credits (pay-as-you-go)
4. Set environment variable: `export AISA_API_KEY="your-key"`

## Full API Reference

See [API Reference](https://aisa.mintlify.app/api-reference/introduction) for complete endpoint documentation.

## Resources

- [AIsa Verity](https://github.com/AIsa-team/verity) - Reference implementation of confidence-scored search agent
- [AIsa Documentation](https://aisa.mintlify.app) - Complete API documentation