Agent Zero - MyFaiss `_faiss_index` AttributeError Fix

# Agent Zero - MyFaiss `_faiss_index` AttributeError Fix

This is a fix for the `AttributeError: 'MyFaiss' object has no attribute '_faiss_index'` error in Agent Zero.

## Problem

When running Agent Zero, you may encounter this error:

```
AttributeError: 'MyFaiss' object has no attribute '_faiss_index'
```

This occurs in the `asimilarity_search_with_score` method when the FAISS vector store is deserialized from disk but the `_faiss_index` attribute isn't properly initialized.

## Root Cause

The `MyFaiss` class inherits from LangChain's `FAISS` class. When loading a saved vector store using `load_local()`, the deserialized object may not have the `_faiss_index` attribute set correctly, causing failures when the code tries to call `self._faiss_index.search()`.

## Solution

Add a safety check before accessing `_faiss_index` that:
1. Checks if the attribute exists
2. Falls back to `self.index` if available
3. Delegates to parent class method as a last resort

## Fixed Code

Replace the `asimilarity_search_with_score` method in `/a0/python/helpers/memory.py`:

**Before (lines ~169-194):**
```python
async def asimilarity_search_with_score(self, query: str, k: int = 4, filter: Any = None, fetch_k: int = 100, **kwargs) -> List[Document]:
    """Override to handle missing documents in docstore (index/docstore desync)."""
    # Get embedding function
    if hasattr(self, '_embeddings'):
        embed_fn = self._embeddings
    elif hasattr(self, 'embedding_function'):
        embed_fn = self.embedding_function
    else:
        return await super().asimilarity_search_with_score(query, k, filter, fetch_k, **kwargs)

    # Embed the query
    vector = await embed_fn.aembed_query(query)

    # Search FAISS index directly
    import numpy as np
    scores, indices = self._faiss_index.search(np.array([vector]), fetch_k)
```

**After (with fix):**
```python
async def asimilarity_search_with_score(self, query: str, k: int = 4, filter: Any = None, fetch_k: int = 100, **kwargs) -> List[Document]:
    """Override to handle missing documents in docstore (index/docstore desync)."""
    # Get embedding function
    if hasattr(self, '_embeddings'):
        embed_fn = self._embeddings
    elif hasattr(self, 'embedding_function'):
        embed_fn = self.embedding_function
    else:
        return await super().asimilarity_search_with_score(query, k, filter, fetch_k, **kwargs)

    # Embed the query
    vector = await embed_fn.aembed_query(query)

    # Ensure _faiss_index is available (fix for deserialization issue)
    if not hasattr(self, '_faiss_index') or self._faiss_index is None:
        if hasattr(self, 'index') and self.index is not None:
            self._faiss_index = self.index
        else:
            return await super().asimilarity_search_with_score(query, k, filter, fetch_k, **kwargs)

    # Search FAISS index directly
    import numpy as np
    scores, indices = self._faiss_index.search(np.array([vector]), fetch_k)
```

## Additional Fixes Required

### 1. Fix `@functools.wraps` decorator (line ~26)

**Before:**
```python
 @functools.wraps(_original_similarity_search_by_vector)
```

**After:**
```python
@functools.wraps(_original_similarity_search_by_vector)
```

(Remove the leading space)

## Deployment Instructions

### Option 1: Manual Edit (Recommended)

```bash
# Access your Agent Zero container
docker exec -it <container-id> bash

# Backup the original file
cp /a0/python/helpers/memory.py /a0/python/helpers/memory.py.bak

# Edit the file
nano /a0/python/helpers/memory.py

# Make the changes described above
# Save with Ctrl+O, Enter, then exit with Ctrl+X

# Restart the container
exit
docker restart <container-id>
```

### Option 2: Automated Python Script

Run this from your host machine:

```bash
docker exec -it <container-id> bash -c "
python3 << 'PYEOF'
file = '/a0/python/helpers/memory.py'
backup = '/a0/python/helpers/memory.py.bak'

# Create backup
with open(file, 'r') as f:
    content = f.read()
with open(backup, 'w') as f:
    f.write(content)

# Fix the decorator (remove leading space)
content = content.replace(' @functools.wraps', '@functools.wraps')

# Fix the asimilarity_search_with_score method
old = '''        # Embed the query
        vector = await embed_fn.aembed_query(query)

        # Search FAISS index directly
        import numpy as np
        scores, indices = self._faiss_index.search(np.array([vector]), fetch_k)'''

new = '''        # Embed the query
        vector = await embed_fn.aembed_query(query)

        # Ensure _faiss_index is available (fix for deserialization issue)
        if not hasattr(self, '_faiss_index') or self._faiss_index is None:
            if hasattr(self, 'index') and self.index is not None:
                self._faiss_index = self.index
            else:
                return await super().asimilarity_search_with_score(query, k, filter, fetch_k, **kwargs)

        # Search FAISS index directly
        import numpy as np
        scores, indices = self._faiss_index.search(np.array([vector]), fetch_k)'''

content = content.replace(old, new)
with open(file, 'w') as f:
    f.write(content)
print('Fix applied successfully!')
PYEOF
"

# Restart container
docker restart <container-id>
```

## Verification

After applying the fix, the error should no longer occur when Agent Zero searches the memory vector store.

## Files Modified

- `/a0/python/helpers/memory.py` - Main fix location

## Compatibility

- Agent Zero (all recent versions)
- Python 3.11+
- LangChain Community FAISS vector store

---

**Note:** Always backup your files before making changes. The backup will be saved as `memory.py.bak` in the same directory.