[BUG] RAG filtering error in types.py: len(where) != 1 fails when where has 2 conditions

[AstrumPush]

Environment

• Provider (select one):

  • Deepseek

• PraisonAI 4.6.46

• praisonai-tools 0.2.30

• praisonaiagents 1.6.46

• Operating System: Windows 11

Thank you for open-sourcing this project! Today I spent a lot of time debugging an issue related to RAG: I found that even though my question was explicitly mentioned in the documentation, the system still couldn't answer it correctly. After debugging, I discovered that the retrieval step successfully found relevant chunks, but an error occurred during the filtering step in types.py (and the error was not printed). Specifically, the where parameter passed to the filter contained 2 conditions, but the code checks for len(where) != 1. I don't understand the design rationale behind this. Also, I'm not sure how to resolve this issue. Could you please provide some guidance on how to get the RAG test to work properly? Thank you very much!

Full Code

from praisonaiagents import Agent

agent = Agent(
name="RAG Agent",
instructions="Answer questions using the knowledge base",
knowledge=["docs/manual.pdf"] # Your documents
)

 results = self.knowledge.search(
            query=query,
            user_id=user_id,
            agent_id=agent_id,
            run_id=run_id,
            **kwargs,
        )

1. pipeline.py
    results = self.knowledge.search(
               query=query,
               user_id=user_id,
               agent_id=agent_id,
               run_id=run_id,
               **kwargs,
           )

2. knowledge.py
   results = self.memory.search(query, user_id=user_id, agent_id=agent_id, run_id=run_id, rerank=rerank, **kwargs)

3. factories.py

4. CollectionCommon.py

5. types.py

[praisonai-triage-agent]

👋 Thanks for opening this issue, @AstrumPush!

A maintainer will review this shortly. In the meantime:

• Make sure you've included steps to reproduce (for bugs)
• Check existing issues for duplicates
• Review the documentation for related guides

A maintainer can trigger deeper analysis by commenting @claude on this issue.

[praisonai-triage-agent]

Claude finished @praisonai-triage-agent[bot]'s task —— View job • claude/issue-1728-20260523-1153 • [Create PR ➔](https://github.com/MervinPraison/PraisonAI/compare/main...claude/issue-1728-20260523-1153?quick_pull=1&title=Issue%20%231728%3A%20Changes%20from%20Claude&body=This%20PR%20addresses%20issue%20%231728%0A%0AGenerated%20with%20%5BClaude%20Code%5D(https%3A%2F%2Fclaude.ai%2Fcode)

---

RAG filtering error fixed! ✅

Todo List:

• Setup git identity and authentication
• Read AGENTS.md for architecture guidelines
• Locate and examine the types.py file mentioned in the issue
• Analyze the where clause filtering logic
• Fix the filtering condition to handle multiple where clauses
• Test the changes
• Commit and push the fix
• Create pull request

Problem Identified:

The issue was in the ChromaKnowledgeAdapter in factories.py:176-177. When multiple filter conditions are present (user_id, agent_id, run_id), the code builds a dictionary with multiple keys, but ChromaDB's internal validation expects exactly one condition (len(where) != 1).

Solution Implemented:

Modified the filtering logic to use ChromaDB's $and operator when multiple conditions are present:

# Add where filter if any conditions specified
if where_filter:
    # ChromaDB validation expects exactly one condition in the where clause
    # When multiple conditions exist, use $and operator to combine them
    if len(where_filter) == 1:
        query_kwargs["where"] = where_filter
    else:
        # Convert multiple conditions to ChromaDB's $and format
        and_conditions = [{k: v} for k, v in where_filter.items()]
        query_kwargs["where"] = {"$and": and_conditions}

Testing:

✅ Created and ran verification test that confirms:

• Single conditions pass through unchanged
• Multiple conditions are properly formatted with $and operator
• Import and basic logic work correctly

Result:

✅ Pull Request Created: #1729

This fix resolves the RAG filtering issue by properly formatting multiple where conditions according to ChromaDB's requirements, while maintaining backward compatibility for single condition filters.

--- | Branch: claude/issue-1728-20260523-1153

[AstrumPush]

I encountered an error in the deduplication function in context.py – specifically in deduplicate_chunks. The error message is:

AttributeError: 'SearchResultItem' object has no attribute 'get'
Code Analysis
The function currently assumes that the results parameter is a list of dictionaries (List[Dict[str, Any]]), so it uses result.get("text") to extract the text. However, the actual input is a list of SearchResultItem objects, which do not have a .get() method, causing the error.

Relevant code snippet:

def deduplicate_chunks(
results: List[Dict[str, Any]],
...
) -> ...:
for result in results:
text = result.get("text") or result.get("memory", "")

Expected Behavior

The function should be able to handle SearchResultItem objects (e.g., via attribute access like result.text), or the function signature and the actual call site need to be aligned.

Request for Guidance

How should I fix this issue? Should I modify deduplicate_chunks to be compatible with SearchResultItem, or adjust the caller to pass dictionaries instead? Any guidance would be appreciated. Thank you!