Bug: joy_trust_tool.py reads trust_score from wrong location in API response

[tlkc888-Jenkins]

Bug Description

The check_trust method in joy_trust_tool.py reads trust fields from the top-level API response instead of from the agent object inside the agents array.

Current Code (Line ~166)

data = response.json()
trust_score = data.get("trust_score", 0.0)  # ❌ Wrong - reads from top level

Actual API Response Format

{
  "agents": [
    {
      "id": "ag_xxx",
      "name": "AgentName",
      "trust_score": 2.3,
      "verified": true,
      ...
    }
  ],
  "count": 1
}

Fix

data = response.json()
agents = data.get("agents", [])

# Find matching agent by name
agent = None
for a in agents:
    if a.get("name", "").lower() == agent_name.lower():
        agent = a
        break

# Fallback to first agent if exact match not found
if not agent and agents:
    agent = agents[0]

if not agent:
    return {
        "agent_name": agent_name,
        "trust_score": 0.0,
        "verified": False,
        "meets_threshold": False,
        "error": f"Agent '{agent_name}' not found on Joy Trust Network"
    }

trust_score = agent.get("trust_score", 0.0)

Impact

Without this fix, check_trust_score() always returns trust_score: 0.0 because it's reading from the wrong location in the response.

Related

• PR feat: add marketplace tools - Pinchwork, AgentID, Joy Trust #15 (merged) - Original implementation
• CodeRabbit review flagged this as Critical

[github-actions]

🤖 Auto-triaging with Claude. Analyzing issue and implementing a fix...

[tlkc888-Jenkins]

@MervinPraison @claude This is a blocking bug - the Joy Trust integration currently always returns trust_score: 0.0 because of the response parsing issue.

Here's the complete fixed check_trust method to replace lines 116-188:

def check_trust(self, agent_name: str, min_score: Optional[float] = None) -> Dict[str, Any]:
    """Check an agent's trust score on Joy Trust Network.
    
    Args:
        agent_name: Name/identifier of the agent to check
        min_score: Minimum acceptable trust score (uses config default if not provided)
    
    Returns:
        Dictionary containing trust information and verification status
    """
    import time
    
    if not agent_name:
        return {
            "agent_name": "",
            "trust_score": 0.0,
            "verified": False,
            "meets_threshold": False,
            "reputation": {},
            "recommendations": 0,
            "error": "agent_name is required"
        }
    
    min_threshold = min_score if min_score is not None else self.config.min_score
    
    # Check cache first
    cache_key = f"{agent_name}_{min_threshold}"
    if cache_key in self._cache:
        cached_result = self._cache[cache_key]
        if cached_result.get('_cached_at', 0) + self.config.cache_duration > time.time():
            logger.debug(f"Using cached trust score for {agent_name}")
            return cached_result
    
    try:
        import httpx
    except ImportError:
        return {
            "agent_name": agent_name,
            "trust_score": 0.0,
            "verified": False,
            "meets_threshold": False,
            "reputation": {},
            "recommendations": 0,
            "error": (
                "httpx is required for Joy Trust Network integration. "
                "Install with: pip install praisonai-tools[marketplace] or pip install httpx"
            )
        }
    
    try:
        with httpx.Client(timeout=self.config.timeout_seconds) as client:
            headers = {}
            if self.api_key:
                headers["x-api-key"] = self.api_key
                
            response = client.get(
                "https://joy-connect.fly.dev/agents/discover",
                params={"query": agent_name},
                headers=headers
            )
            response.raise_for_status()
            
            data = response.json()
            
            # FIX: Extract agent from the agents array, not top level
            agents = data.get("agents", [])
            
            # Find matching agent by name (case-insensitive)
            agent = None
            for a in agents:
                if a.get("name", "").lower() == agent_name.lower():
                    agent = a
                    break
            
            # Fallback to first agent if no exact match
            if not agent and agents:
                agent = agents[0]
            
            if not agent:
                return {
                    "agent_name": agent_name,
                    "trust_score": 0.0,
                    "verified": False,
                    "meets_threshold": False,
                    "reputation": {},
                    "recommendations": 0,
                    "error": f"Agent '{agent_name}' not found on Joy Trust Network"
                }
            
            # Now read from the agent object
            trust_score = agent.get("trust_score", 0.0)
            
            result = {
                "agent_name": agent.get("name", agent_name),
                "agent_id": agent.get("id"),
                "trust_score": trust_score,
                "verified": agent.get("verified", False),
                "meets_threshold": trust_score >= min_threshold,
                "threshold_used": min_threshold,
                "vouch_count": agent.get("vouch_count", 0),
                "capabilities": agent.get("capabilities", []),
                "tier": agent.get("tier", "free"),
                "badges": agent.get("badges", []),
                "error": None,
                "_cached_at": time.time()
            }
            
            # Cache the result
            self._cache[cache_key] = result
            
            return result
            
    except httpx.RequestError as e:
        logger.error(f"Joy Trust request error: {e}")
        return {
            "agent_name": agent_name,
            "trust_score": 0.0,
            "verified": False,
            "meets_threshold": self.config.fallback_on_error,
            "error": f"Connection error: {e}",
            "fallback_used": self.config.fallback_on_error
        }
    except Exception as e:
        logger.error(f"Joy Trust unexpected error: {e}")
        return {
            "agent_name": agent_name,
            "trust_score": 0.0,
            "verified": False,
            "meets_threshold": self.config.fallback_on_error,
            "error": f"Unexpected error: {e}",
            "fallback_used": self.config.fallback_on_error
        }

Key changes:

1. Parse agents array from response
2. Find matching agent by name
3. Read trust_score from agent object, not top level
4. Use query param instead of name param (that's what the API expects)
5. Pass API key via header, not query param
6. Return more useful fields from the agent object

[tlkc888-Jenkins]

Bump - this is blocking Joy Trust integration from working. The check_trust_score() function always returns trust_score: 0.0 because it reads from the wrong level in the API response.

Happy to submit a PR directly if that's easier.