swarms/bug_report.md

Bug Report: Swarms Codebase Issues

Bug 1: Error Handling in Daemon Thread (Critical)

Location: swarms/structs/agent.py lines 1446-1453

Description: The _handle_run_error method creates a daemon thread to handle errors, but the exception raised in __handle_run_error will not propagate back to the main thread. This causes silent failures where errors are logged but not properly propagated to the caller.

Type: Concurrency/Error Handling Bug

Severity: Critical - Can lead to silent failures

Current Code:

def _handle_run_error(self, error: any):
    process_thread = threading.Thread(
        target=self.__handle_run_error,
        args=(error,),
        daemon=True,
    )
    process_thread.start()

Problem:

• The daemon thread will exit when the main thread exits
• The raise error at the end of __handle_run_error occurs in the daemon thread, not the main thread
• This means exceptions are lost and not properly handled by the calling code

Fix: Remove the threading wrapper and call the error handler directly, or use proper exception propagation.

---

Bug 2: Method Name Typo (Logic Error)

Location: swarms/structs/agent.py lines 2128 and 2122

Description: There are two related typos in the response filtering functionality:

1. The method apply_reponse_filters has a typo in the name - it should be apply_response_filters
2. The add_response_filter method accesses self.reponse_filters instead of self.response_filters

Type: Naming/Logic Error

Severity: Medium - Can cause AttributeError when called

Current Code:

def add_response_filter(self, filter_word: str) -> None:
    logger.info(f"Adding response filter: {filter_word}")
    self.reponse_filters.append(filter_word)  # TYPO: reponse_filters

def apply_reponse_filters(self, response: str) -> str:  # TYPO: apply_reponse_filters
    """
    Apply the response filters to the response
    """
    logger.info(
        f"Applying response filters to response: {response}"
    )
    for word in self.response_filters:
        response = response.replace(word, "[FILTERED]")
    return response

Problem:

• Method name is misspelled: apply_reponse_filters instead of apply_response_filters
• Attribute access is misspelled: self.reponse_filters instead of self.response_filters
• The method is called correctly in filtered_run method, suggesting these are typos

Fix: Fix both typos to use correct spelling.

---

Bug 3: Document Ingestion Logic Error (Data Loss)

Location: swarms/structs/agent.py lines 2193-2212

Description: The ingest_docs method has a logic error where it processes all documents in a loop but only retains the data from the last document. All previous documents are processed but their data is overwritten and lost.

Type: Logic Error

Severity: High - Causes data loss

Current Code:

def ingest_docs(self, docs: List[str], *args, **kwargs):
    """Ingest the docs into the memory

    Args:
        docs (List[str]): Documents of pdfs, text, csvs

    Returns:
        None
    """
    try:
        for doc in docs:
            data = data_to_text(doc)

        return self.short_memory.add(
            role=self.user_name, content=data
        )
    except Exception as error:
        logger.info(f"Error ingesting docs: {error}", "red")

Problem:

• The data variable is overwritten on each iteration
• Only the last document's data is actually added to memory
• All previous documents are processed but their data is lost
• The method should either process documents individually or combine all data

Fix: Accumulate all document data or process each document individually.

---

Impact Assessment

1. Bug 1 (Critical): Can cause silent failures in production, making debugging difficult
2. Bug 2 (Medium): Will cause AttributeError when the method is called correctly
3. Bug 3 (High): Causes data loss when ingesting multiple documents

Fixes Applied

Bug 1 Fix - Error Handling

Status: ✅ FIXED

Changed the _handle_run_error method to call __handle_run_error directly instead of using a daemon thread:

def _handle_run_error(self, error: any):
    # Handle error directly instead of using daemon thread
    # to ensure proper exception propagation
    self.__handle_run_error(error)

Bug 2 Fix - Method Name Typos

Status: ✅ FIXED

Fixed both typos in the response filtering functionality:

1. Renamed apply_reponse_filters to apply_response_filters
2. Fixed self.reponse_filters to self.response_filters

Bug 3 Fix - Document Ingestion Logic

Status: ✅ FIXED

Modified the ingest_docs method to process all documents and combine their content:

def ingest_docs(self, docs: List[str], *args, **kwargs):
    try:
        # Process all documents and combine their content
        all_data = []
        for doc in docs:
            data = data_to_text(doc)
            all_data.append(f"Document: {doc}\n{data}")
        
        # Combine all document content
        combined_data = "\n\n".join(all_data)

        return self.short_memory.add(
            role=self.user_name, content=combined_data
        )
    except Exception as error:
        logger.info(f"Error ingesting docs: {error}", "red")

Recommendations

1. ✅ Fixed the error handling to properly propagate exceptions
2. ✅ Corrected the method name typos
3. ✅ Fixed the document ingestion logic to process all documents
4. Add unit tests to prevent similar issues in the future
5. Consider adding linting rules to catch method name typos
6. Consider code review processes to catch similar issues