516f2963e0
* Adapt to turbopuffer embedder * Make turbopuffer search more efficient over all source ids * Combine turbopuffer and pinecone hybrid * Fix test sources
852 lines
37 KiB
Python
852 lines
37 KiB
Python
import asyncio
|
|
import re
|
|
from typing import Any, Dict, List, Optional
|
|
|
|
from letta.constants import PINECONE_TEXT_FIELD_NAME
|
|
from letta.functions.types import FileOpenRequest
|
|
from letta.helpers.pinecone_utils import search_pinecone_index, should_use_pinecone
|
|
from letta.helpers.tpuf_client import should_use_tpuf
|
|
from letta.log import get_logger
|
|
from letta.otel.tracing import trace_method
|
|
from letta.schemas.agent import AgentState
|
|
from letta.schemas.enums import VectorDBProvider
|
|
from letta.schemas.sandbox_config import SandboxConfig
|
|
from letta.schemas.source import Source
|
|
from letta.schemas.tool import Tool
|
|
from letta.schemas.tool_execution_result import ToolExecutionResult
|
|
from letta.schemas.user import User
|
|
from letta.services.agent_manager import AgentManager
|
|
from letta.services.block_manager import BlockManager
|
|
from letta.services.file_manager import FileManager
|
|
from letta.services.file_processor.chunker.line_chunker import LineChunker
|
|
from letta.services.files_agents_manager import FileAgentManager
|
|
from letta.services.job_manager import JobManager
|
|
from letta.services.message_manager import MessageManager
|
|
from letta.services.passage_manager import PassageManager
|
|
from letta.services.source_manager import SourceManager
|
|
from letta.services.tool_executor.tool_executor_base import ToolExecutor
|
|
from letta.utils import get_friendly_error_msg
|
|
|
|
|
|
class LettaFileToolExecutor(ToolExecutor):
|
|
"""Executor for Letta file tools with direct implementation of functions."""
|
|
|
|
# Production safety constants
|
|
MAX_FILE_SIZE_BYTES = 50 * 1024 * 1024 # 50MB limit per file
|
|
MAX_TOTAL_CONTENT_SIZE = 200 * 1024 * 1024 # 200MB total across all files
|
|
MAX_REGEX_COMPLEXITY = 1000 # Prevent catastrophic backtracking
|
|
MAX_MATCHES_PER_FILE = 20 # Limit matches per file (legacy, not used with new pagination)
|
|
MAX_TOTAL_MATCHES = 50 # Keep original value for semantic search
|
|
GREP_PAGE_SIZE = 20 # Number of grep matches to show per page
|
|
GREP_TIMEOUT_SECONDS = 30 # Max time for grep_files operation
|
|
MAX_CONTEXT_LINES = 1 # Lines of context around matches
|
|
MAX_TOTAL_COLLECTED = 1000 # Reasonable upper limit to prevent memory issues
|
|
|
|
def __init__(
|
|
self,
|
|
message_manager: MessageManager,
|
|
agent_manager: AgentManager,
|
|
block_manager: BlockManager,
|
|
job_manager: JobManager,
|
|
passage_manager: PassageManager,
|
|
actor: User,
|
|
):
|
|
super().__init__(
|
|
message_manager=message_manager,
|
|
agent_manager=agent_manager,
|
|
block_manager=block_manager,
|
|
job_manager=job_manager,
|
|
passage_manager=passage_manager,
|
|
actor=actor,
|
|
)
|
|
|
|
# TODO: This should be passed in to for testing purposes
|
|
self.files_agents_manager = FileAgentManager()
|
|
self.file_manager = FileManager()
|
|
self.source_manager = SourceManager()
|
|
self.logger = get_logger(__name__)
|
|
|
|
async def execute(
|
|
self,
|
|
function_name: str,
|
|
function_args: dict,
|
|
tool: Tool,
|
|
actor: User,
|
|
agent_state: Optional[AgentState] = None,
|
|
sandbox_config: Optional[SandboxConfig] = None,
|
|
sandbox_env_vars: Optional[Dict[str, Any]] = None,
|
|
) -> ToolExecutionResult:
|
|
if agent_state is None:
|
|
raise ValueError("Agent state is required for file tools")
|
|
|
|
function_map = {
|
|
"open_files": self.open_files,
|
|
"grep_files": self.grep_files,
|
|
"semantic_search_files": self.semantic_search_files,
|
|
}
|
|
|
|
if function_name not in function_map:
|
|
raise ValueError(f"Unknown function: {function_name}")
|
|
|
|
function_args_copy = function_args.copy()
|
|
try:
|
|
func_return = await function_map[function_name](agent_state, **function_args_copy)
|
|
return ToolExecutionResult(
|
|
status="success",
|
|
func_return=func_return,
|
|
agent_state=agent_state,
|
|
)
|
|
except Exception as e:
|
|
return ToolExecutionResult(
|
|
status="error",
|
|
func_return=e,
|
|
agent_state=agent_state,
|
|
stderr=[get_friendly_error_msg(function_name=function_name, exception_name=type(e).__name__, exception_message=str(e))],
|
|
)
|
|
|
|
@trace_method
|
|
async def open_files(self, agent_state: AgentState, file_requests: List[FileOpenRequest], close_all_others: bool = False) -> str:
|
|
"""Open one or more files and load their contents into memory blocks."""
|
|
# Parse raw dictionaries into FileOpenRequest objects if needed
|
|
parsed_requests = []
|
|
for req in file_requests:
|
|
if isinstance(req, dict):
|
|
# LLM returned a dictionary, parse it into FileOpenRequest
|
|
parsed_requests.append(FileOpenRequest(**req))
|
|
elif isinstance(req, FileOpenRequest):
|
|
# Already a FileOpenRequest object
|
|
parsed_requests.append(req)
|
|
else:
|
|
raise ValueError(f"Invalid file request type: {type(req)}. Expected dict or FileOpenRequest.")
|
|
|
|
file_requests = parsed_requests
|
|
|
|
# Validate file count first
|
|
if len(file_requests) > agent_state.max_files_open:
|
|
raise ValueError(
|
|
f"Cannot open {len(file_requests)} files: exceeds configured maximum limit of {agent_state.max_files_open} files"
|
|
)
|
|
|
|
if not file_requests:
|
|
raise ValueError("No file requests provided")
|
|
|
|
# Extract file names for various operations
|
|
file_names = [req.file_name for req in file_requests]
|
|
|
|
# Get all currently attached files for error reporting
|
|
file_blocks = agent_state.memory.file_blocks
|
|
attached_file_names = [fb.label for fb in file_blocks]
|
|
|
|
# Close all other files if requested
|
|
closed_by_close_all_others = []
|
|
if close_all_others:
|
|
closed_by_close_all_others = await self.files_agents_manager.close_all_other_files(
|
|
agent_id=agent_state.id, keep_file_names=file_names, actor=self.actor
|
|
)
|
|
|
|
# Process each file
|
|
opened_files = []
|
|
all_closed_files = []
|
|
all_previous_ranges = {} # Collect all previous ranges from all files
|
|
|
|
for file_request in file_requests:
|
|
file_name = file_request.file_name
|
|
offset = file_request.offset
|
|
length = file_request.length
|
|
|
|
# Use 0-indexed offset/length directly for LineChunker
|
|
start, end = None, None
|
|
if offset is not None or length is not None:
|
|
if offset is not None and offset < 0:
|
|
raise ValueError(f"Offset for file {file_name} must be >= 0 (0-indexed), got {offset}")
|
|
if length is not None and length < 1:
|
|
raise ValueError(f"Length for file {file_name} must be >= 1, got {length}")
|
|
|
|
# Use offset directly as it's already 0-indexed
|
|
start = offset if offset is not None else None
|
|
if start is not None and length is not None:
|
|
end = start + length
|
|
else:
|
|
end = None
|
|
|
|
# Validate file exists and is attached to agent
|
|
file_agent = await self.files_agents_manager.get_file_agent_by_file_name(
|
|
agent_id=agent_state.id, file_name=file_name, actor=self.actor
|
|
)
|
|
|
|
if not file_agent:
|
|
raise ValueError(
|
|
f"{file_name} not attached - did you get the filename correct? Currently you have the following files attached: {attached_file_names}"
|
|
)
|
|
|
|
file_id = file_agent.file_id
|
|
file = await self.file_manager.get_file_by_id(file_id=file_id, actor=self.actor, include_content=True)
|
|
|
|
# Process file content
|
|
content_lines = LineChunker().chunk_text(file_metadata=file, start=start, end=end, validate_range=True)
|
|
visible_content = "\n".join(content_lines)
|
|
|
|
# Handle LRU eviction and file opening
|
|
closed_files, was_already_open, previous_ranges = await self.files_agents_manager.enforce_max_open_files_and_open(
|
|
agent_id=agent_state.id,
|
|
file_id=file_id,
|
|
file_name=file_name,
|
|
source_id=file.source_id,
|
|
actor=self.actor,
|
|
visible_content=visible_content,
|
|
max_files_open=agent_state.max_files_open,
|
|
start_line=start + 1 if start is not None else None, # convert to 1-indexed for user display
|
|
end_line=end if end is not None else None, # end is already exclusive, shows as 1-indexed inclusive
|
|
)
|
|
|
|
opened_files.append(file_name)
|
|
all_closed_files.extend(closed_files)
|
|
all_previous_ranges.update(previous_ranges) # Merge previous ranges from this file
|
|
|
|
# Update access timestamps for all opened files efficiently
|
|
await self.files_agents_manager.mark_access_bulk(agent_id=agent_state.id, file_names=file_names, actor=self.actor)
|
|
|
|
# Helper function to format previous range info
|
|
def format_previous_range(file_name: str) -> str:
|
|
if file_name in all_previous_ranges:
|
|
old_start, old_end = all_previous_ranges[file_name]
|
|
if old_start is not None and old_end is not None:
|
|
return f" (previously lines {old_start}-{old_end})"
|
|
elif old_start is not None:
|
|
return f" (previously lines {old_start}-end)"
|
|
else:
|
|
return " (previously full file)"
|
|
return ""
|
|
|
|
# Build unified success message - treat single and multiple files consistently
|
|
file_summaries = []
|
|
for req in file_requests:
|
|
previous_info = format_previous_range(req.file_name)
|
|
if req.offset is not None and req.length is not None:
|
|
# Display as 1-indexed for user readability: (offset+1) to (offset+length)
|
|
start_line = req.offset + 1
|
|
end_line = req.offset + req.length
|
|
file_summaries.append(f"{req.file_name} (lines {start_line}-{end_line}){previous_info}")
|
|
elif req.offset is not None:
|
|
# Display as 1-indexed
|
|
start_line = req.offset + 1
|
|
file_summaries.append(f"{req.file_name} (lines {start_line}-end){previous_info}")
|
|
else:
|
|
file_summaries.append(f"{req.file_name}{previous_info}")
|
|
|
|
if len(file_requests) == 1:
|
|
success_msg = f"* Opened {file_summaries[0]}"
|
|
else:
|
|
success_msg = f"* Opened {len(file_requests)} files: {', '.join(file_summaries)}"
|
|
|
|
# Add information about closed files
|
|
if closed_by_close_all_others:
|
|
success_msg += f"\nNote: Closed {len(closed_by_close_all_others)} file(s) due to close_all_others=True: {', '.join(closed_by_close_all_others)}"
|
|
|
|
if all_closed_files:
|
|
success_msg += (
|
|
f"\nNote: Closed {len(all_closed_files)} least recently used file(s) due to open file limit: {', '.join(all_closed_files)}"
|
|
)
|
|
|
|
return success_msg
|
|
|
|
def _validate_regex_pattern(self, pattern: str) -> None:
|
|
"""Validate regex pattern to prevent catastrophic backtracking."""
|
|
if len(pattern) > self.MAX_REGEX_COMPLEXITY:
|
|
raise ValueError(f"Pattern too complex: {len(pattern)} chars > {self.MAX_REGEX_COMPLEXITY} limit")
|
|
|
|
# Test compile the pattern to catch syntax errors early
|
|
try:
|
|
re.compile(pattern, re.IGNORECASE | re.MULTILINE)
|
|
except re.error as e:
|
|
raise ValueError(f"Invalid regex pattern: {e}")
|
|
|
|
def _get_context_lines(
|
|
self,
|
|
formatted_lines: List[str],
|
|
match_line_num: int,
|
|
context_lines: int,
|
|
) -> List[str]:
|
|
"""Get context lines around a match from already-chunked lines.
|
|
|
|
Args:
|
|
formatted_lines: Already chunked lines from LineChunker (format: "line_num: content")
|
|
match_line_num: The 1-based line number of the match
|
|
context_lines: Number of context lines before and after
|
|
"""
|
|
if not formatted_lines or context_lines < 0:
|
|
return []
|
|
|
|
# Find the index of the matching line in the formatted_lines list
|
|
match_formatted_idx = None
|
|
for i, line in enumerate(formatted_lines):
|
|
if line and ":" in line:
|
|
try:
|
|
line_num = int(line.split(":", 1)[0].strip())
|
|
if line_num == match_line_num:
|
|
match_formatted_idx = i
|
|
break
|
|
except ValueError:
|
|
continue
|
|
|
|
if match_formatted_idx is None:
|
|
return []
|
|
|
|
# Calculate context range with bounds checking
|
|
start_idx = max(0, match_formatted_idx - context_lines)
|
|
end_idx = min(len(formatted_lines), match_formatted_idx + context_lines + 1)
|
|
|
|
# Extract context lines and add match indicator
|
|
context_lines_with_indicator = []
|
|
for i in range(start_idx, end_idx):
|
|
line = formatted_lines[i]
|
|
prefix = ">" if i == match_formatted_idx else " "
|
|
context_lines_with_indicator.append(f"{prefix} {line}")
|
|
|
|
return context_lines_with_indicator
|
|
|
|
@trace_method
|
|
async def grep_files(
|
|
self,
|
|
agent_state: AgentState,
|
|
pattern: str,
|
|
include: Optional[str] = None,
|
|
context_lines: Optional[int] = 1,
|
|
offset: Optional[int] = None,
|
|
) -> str:
|
|
"""
|
|
Search for pattern in all attached files and return matches with context.
|
|
|
|
Args:
|
|
agent_state: Current agent state
|
|
pattern: Regular expression pattern to search for
|
|
include: Optional pattern to filter filenames to include in the search
|
|
context_lines (Optional[int]): Number of lines of context to show before and after each match.
|
|
Equivalent to `-C` in grep_files. Defaults to 1.
|
|
offset (Optional[int]): Number of matches to skip before showing results. Used for pagination.
|
|
Defaults to 0 (show from first match).
|
|
|
|
Returns:
|
|
Formatted string with search results, file names, line numbers, and context
|
|
"""
|
|
if not pattern or not pattern.strip():
|
|
raise ValueError("Empty search pattern provided")
|
|
|
|
pattern = pattern.strip()
|
|
self._validate_regex_pattern(pattern)
|
|
|
|
# Validate include pattern if provided
|
|
include_regex = None
|
|
if include and include.strip():
|
|
include = include.strip()
|
|
# Convert glob pattern to regex if it looks like a glob pattern
|
|
if "*" in include and not any(c in include for c in ["^", "$", "(", ")", "[", "]", "{", "}", "\\", "+"]):
|
|
# Simple glob to regex conversion
|
|
include_pattern = include.replace(".", r"\.").replace("*", ".*").replace("?", ".")
|
|
if not include_pattern.endswith("$"):
|
|
include_pattern += "$"
|
|
else:
|
|
include_pattern = include
|
|
|
|
self._validate_regex_pattern(include_pattern)
|
|
include_regex = re.compile(include_pattern, re.IGNORECASE)
|
|
|
|
# Get all attached files for this agent
|
|
file_agents = await self.files_agents_manager.list_files_for_agent(
|
|
agent_id=agent_state.id, per_file_view_window_char_limit=agent_state.per_file_view_window_char_limit, actor=self.actor
|
|
)
|
|
|
|
if not file_agents:
|
|
return "No files are currently attached to search"
|
|
|
|
# Filter files by filename pattern if include is specified
|
|
if include_regex:
|
|
original_count = len(file_agents)
|
|
file_agents = [fa for fa in file_agents if include_regex.search(fa.file_name)]
|
|
if not file_agents:
|
|
return f"No files match the filename pattern '{include}' (filtered {original_count} files)"
|
|
|
|
# Validate offset parameter
|
|
if offset is not None and offset < 0:
|
|
offset = 0 # Treat negative offsets as 0
|
|
|
|
# Compile regex pattern with appropriate flags
|
|
regex_flags = re.MULTILINE
|
|
regex_flags |= re.IGNORECASE
|
|
|
|
pattern_regex = re.compile(pattern, regex_flags)
|
|
|
|
# Collect all matches first (up to a reasonable limit)
|
|
all_matches = [] # List of tuples: (file_name, line_num, context_lines)
|
|
total_content_size = 0
|
|
files_processed = 0
|
|
files_skipped = 0
|
|
files_with_matches = set() # Track files that had matches for LRU policy
|
|
|
|
# Use asyncio timeout to prevent hanging
|
|
async def _search_files():
|
|
nonlocal all_matches, total_content_size, files_processed, files_skipped, files_with_matches
|
|
|
|
for file_agent in file_agents:
|
|
# Load file content
|
|
file = await self.file_manager.get_file_by_id(file_id=file_agent.file_id, actor=self.actor, include_content=True)
|
|
|
|
if not file or not file.content:
|
|
files_skipped += 1
|
|
self.logger.warning(f"Grep: Skipping file {file_agent.file_name} - no content available")
|
|
continue
|
|
|
|
# Check individual file size
|
|
content_size = len(file.content.encode("utf-8"))
|
|
if content_size > self.MAX_FILE_SIZE_BYTES:
|
|
files_skipped += 1
|
|
self.logger.warning(
|
|
f"Grep: Skipping file {file.file_name} - too large ({content_size:,} bytes > {self.MAX_FILE_SIZE_BYTES:,} limit)"
|
|
)
|
|
continue
|
|
|
|
# Check total content size across all files
|
|
total_content_size += content_size
|
|
if total_content_size > self.MAX_TOTAL_CONTENT_SIZE:
|
|
files_skipped += 1
|
|
self.logger.warning(
|
|
f"Grep: Skipping file {file.file_name} - total content size limit exceeded ({total_content_size:,} bytes > {self.MAX_TOTAL_CONTENT_SIZE:,} limit)"
|
|
)
|
|
break
|
|
|
|
files_processed += 1
|
|
|
|
# Use LineChunker to get all lines with proper formatting
|
|
chunker = LineChunker()
|
|
formatted_lines = chunker.chunk_text(file_metadata=file)
|
|
|
|
# Remove metadata header
|
|
if formatted_lines and formatted_lines[0].startswith("[Viewing"):
|
|
formatted_lines = formatted_lines[1:]
|
|
|
|
# Search for matches in formatted lines
|
|
for formatted_line in formatted_lines:
|
|
if len(all_matches) >= self.MAX_TOTAL_COLLECTED:
|
|
# Stop collecting if we hit the upper limit
|
|
break
|
|
|
|
# Extract line number and content from formatted line
|
|
if ":" in formatted_line:
|
|
try:
|
|
line_parts = formatted_line.split(":", 1)
|
|
line_num = int(line_parts[0].strip())
|
|
line_content = line_parts[1].strip() if len(line_parts) > 1 else ""
|
|
except (ValueError, IndexError):
|
|
continue
|
|
|
|
if pattern_regex.search(line_content):
|
|
# Mark this file as having matches for LRU tracking
|
|
files_with_matches.add(file.file_name)
|
|
context = self._get_context_lines(formatted_lines, match_line_num=line_num, context_lines=context_lines or 0)
|
|
|
|
# Store match data for later pagination
|
|
all_matches.append((file.file_name, line_num, context))
|
|
|
|
# Break if we've collected enough matches
|
|
if len(all_matches) >= self.MAX_TOTAL_COLLECTED:
|
|
break
|
|
|
|
# Execute with timeout
|
|
await asyncio.wait_for(_search_files(), timeout=self.GREP_TIMEOUT_SECONDS)
|
|
|
|
# Mark access for files that had matches
|
|
if files_with_matches:
|
|
await self.files_agents_manager.mark_access_bulk(agent_id=agent_state.id, file_names=list(files_with_matches), actor=self.actor)
|
|
|
|
# Handle no matches case
|
|
total_matches = len(all_matches)
|
|
if total_matches == 0:
|
|
summary = f"No matches found for pattern: '{pattern}'"
|
|
if include:
|
|
summary += f" in files matching '{include}'"
|
|
if files_skipped > 0:
|
|
summary += f" (searched {files_processed} files, skipped {files_skipped})"
|
|
return summary
|
|
|
|
# Apply pagination
|
|
start_idx = offset if offset else 0
|
|
end_idx = start_idx + self.GREP_PAGE_SIZE
|
|
paginated_matches = all_matches[start_idx:end_idx]
|
|
|
|
# Check if we hit the collection limit
|
|
hit_collection_limit = len(all_matches) >= self.MAX_TOTAL_COLLECTED
|
|
|
|
# Format the paginated results
|
|
results = []
|
|
|
|
# Build summary showing the range of matches displayed
|
|
if hit_collection_limit:
|
|
# We collected MAX_TOTAL_COLLECTED but there might be more
|
|
summary = f"Found {self.MAX_TOTAL_COLLECTED}+ total matches across {len(files_with_matches)} files (showing matches {start_idx + 1}-{min(end_idx, total_matches)} of {self.MAX_TOTAL_COLLECTED}+)"
|
|
else:
|
|
# We found all matches
|
|
summary = f"Found {total_matches} total matches across {len(files_with_matches)} files (showing matches {start_idx + 1}-{min(end_idx, total_matches)} of {total_matches})"
|
|
|
|
if files_skipped > 0:
|
|
summary += f"\nNote: Skipped {files_skipped} files due to size limits"
|
|
|
|
results.append(summary)
|
|
results.append("=" * 80)
|
|
|
|
# Add file summary - count matches per file
|
|
file_match_counts = {}
|
|
for file_name, _, _ in all_matches:
|
|
file_match_counts[file_name] = file_match_counts.get(file_name, 0) + 1
|
|
|
|
# Sort files by match count (descending) for better overview
|
|
sorted_files = sorted(file_match_counts.items(), key=lambda x: x[1], reverse=True)
|
|
|
|
results.append("\nFiles with matches:")
|
|
for file_name, count in sorted_files:
|
|
if hit_collection_limit and count >= self.MAX_TOTAL_COLLECTED:
|
|
results.append(f" - {file_name}: {count}+ matches")
|
|
else:
|
|
results.append(f" - {file_name}: {count} matches")
|
|
results.append("") # blank line before matches
|
|
|
|
# Format each match in the current page
|
|
for file_name, line_num, context_lines in paginated_matches:
|
|
match_header = f"\n=== {file_name}:{line_num} ==="
|
|
match_content = "\n".join(context_lines)
|
|
results.append(f"{match_header}\n{match_content}")
|
|
|
|
# Add navigation hint
|
|
results.append("") # blank line
|
|
if end_idx < total_matches:
|
|
if hit_collection_limit:
|
|
results.append(f'To see more matches, call: grep_files(pattern="{pattern}", offset={end_idx})')
|
|
results.append(
|
|
f"Note: Only the first {self.MAX_TOTAL_COLLECTED} matches were collected. There may be more matches beyond this limit."
|
|
)
|
|
else:
|
|
results.append(f'To see more matches, call: grep_files(pattern="{pattern}", offset={end_idx})')
|
|
else:
|
|
if hit_collection_limit:
|
|
results.append("Showing last page of collected matches. There may be more matches beyond the collection limit.")
|
|
else:
|
|
results.append("No more matches to show.")
|
|
|
|
return "\n".join(results)
|
|
|
|
@trace_method
|
|
async def semantic_search_files(self, agent_state: AgentState, query: str, limit: int = 5) -> str:
|
|
"""
|
|
Search for text within attached files using semantic search and return passages with their source filenames.
|
|
Uses Pinecone if configured, otherwise falls back to traditional search.
|
|
|
|
Args:
|
|
agent_state: Current agent state
|
|
query: Search query for semantic matching
|
|
limit: Maximum number of results to return (default: 5)
|
|
|
|
Returns:
|
|
Formatted string with search results in IDE/terminal style
|
|
"""
|
|
if not query or not query.strip():
|
|
raise ValueError("Empty search query provided")
|
|
|
|
query = query.strip()
|
|
|
|
# Apply reasonable limit
|
|
limit = min(limit, self.MAX_TOTAL_MATCHES)
|
|
|
|
self.logger.info(f"Semantic search started for agent {agent_state.id} with query '{query}' (limit: {limit})")
|
|
|
|
# Check which vector DB to use - Turbopuffer takes precedence
|
|
attached_sources = await self.agent_manager.list_attached_sources_async(agent_id=agent_state.id, actor=self.actor)
|
|
attached_tpuf_sources = [source for source in attached_sources if source.vector_db_provider == VectorDBProvider.TPUF]
|
|
attached_pinecone_sources = [source for source in attached_sources if source.vector_db_provider == VectorDBProvider.PINECONE]
|
|
|
|
if not attached_tpuf_sources and not attached_pinecone_sources:
|
|
return await self._search_files_native(agent_state, query, limit)
|
|
|
|
results = []
|
|
|
|
# If both have items, we half the limit roughly
|
|
# TODO: This is very hacky bc it skips the re-ranking - but this is a temporary stopgap while we think about migrating data
|
|
|
|
if attached_tpuf_sources and attached_pinecone_sources:
|
|
limit = max(limit // 2, 1)
|
|
|
|
if should_use_tpuf() and attached_tpuf_sources:
|
|
tpuf_result = await self._search_files_turbopuffer(agent_state, attached_tpuf_sources, query, limit)
|
|
results.append(tpuf_result)
|
|
|
|
if should_use_pinecone() and attached_pinecone_sources:
|
|
pinecone_result = await self._search_files_pinecone(agent_state, attached_pinecone_sources, query, limit)
|
|
results.append(pinecone_result)
|
|
|
|
# combine results from both sources
|
|
if results:
|
|
return "\n\n".join(results)
|
|
|
|
# fallback if no results from either source
|
|
return "No results found"
|
|
|
|
async def _search_files_turbopuffer(self, agent_state: AgentState, attached_sources: List[Source], query: str, limit: int) -> str:
|
|
"""Search files using Turbopuffer vector database."""
|
|
|
|
# Get attached sources
|
|
source_ids = [source.id for source in attached_sources]
|
|
if not source_ids:
|
|
return "No valid source IDs found for attached files"
|
|
|
|
# Get all attached files for this agent
|
|
file_agents = await self.files_agents_manager.list_files_for_agent(
|
|
agent_id=agent_state.id, per_file_view_window_char_limit=agent_state.per_file_view_window_char_limit, actor=self.actor
|
|
)
|
|
if not file_agents:
|
|
return "No files are currently attached to search"
|
|
|
|
# Create a map of file_id to file_name for quick lookup
|
|
file_map = {fa.file_id: fa.file_name for fa in file_agents}
|
|
|
|
results = []
|
|
total_hits = 0
|
|
files_with_matches = {}
|
|
|
|
try:
|
|
from letta.helpers.tpuf_client import TurbopufferClient
|
|
|
|
tpuf_client = TurbopufferClient()
|
|
|
|
# Query Turbopuffer for all sources at once
|
|
search_results = await tpuf_client.query_file_passages(
|
|
source_ids=source_ids, # pass all source_ids as a list
|
|
organization_id=self.actor.organization_id,
|
|
actor=self.actor,
|
|
query_text=query,
|
|
search_mode="hybrid", # use hybrid search for best results
|
|
top_k=limit,
|
|
)
|
|
|
|
# Process search results
|
|
for passage, score, metadata in search_results:
|
|
if total_hits >= limit:
|
|
break
|
|
|
|
total_hits += 1
|
|
|
|
# get file name from our map
|
|
file_name = file_map.get(passage.file_id, "Unknown File")
|
|
|
|
# group by file name
|
|
if file_name not in files_with_matches:
|
|
files_with_matches[file_name] = []
|
|
files_with_matches[file_name].append({"text": passage.text, "score": score, "passage_id": passage.id})
|
|
|
|
except Exception as e:
|
|
self.logger.error(f"Turbopuffer search failed: {str(e)}")
|
|
raise e
|
|
|
|
if not files_with_matches:
|
|
return f"No semantic matches found in Turbopuffer for query: '{query}'"
|
|
|
|
# Format results
|
|
passage_num = 0
|
|
for file_name, matches in files_with_matches.items():
|
|
for match in matches:
|
|
passage_num += 1
|
|
|
|
# format each passage with terminal-style header
|
|
score_display = f"(score: {match['score']:.3f})"
|
|
passage_header = f"\n=== {file_name} (passage #{passage_num}) {score_display} ==="
|
|
|
|
# format the passage text
|
|
passage_text = match["text"].strip()
|
|
lines = passage_text.splitlines()
|
|
formatted_lines = []
|
|
for line in lines[:20]: # limit to first 20 lines per passage
|
|
formatted_lines.append(f" {line}")
|
|
|
|
if len(lines) > 20:
|
|
formatted_lines.append(f" ... [truncated {len(lines) - 20} more lines]")
|
|
|
|
passage_content = "\n".join(formatted_lines)
|
|
results.append(f"{passage_header}\n{passage_content}")
|
|
|
|
# mark access for files that had matches
|
|
if files_with_matches:
|
|
matched_file_names = [name for name in files_with_matches.keys() if name != "Unknown File"]
|
|
if matched_file_names:
|
|
await self.files_agents_manager.mark_access_bulk(agent_id=agent_state.id, file_names=matched_file_names, actor=self.actor)
|
|
|
|
# create summary header
|
|
file_count = len(files_with_matches)
|
|
summary = f"Found {total_hits} Turbopuffer matches in {file_count} file{'s' if file_count != 1 else ''} for query: '{query}'"
|
|
|
|
# combine all results
|
|
formatted_results = [summary, "=" * len(summary)] + results
|
|
|
|
self.logger.info(f"Turbopuffer search completed: {total_hits} matches across {file_count} files")
|
|
return "\n".join(formatted_results)
|
|
|
|
async def _search_files_pinecone(self, agent_state: AgentState, attached_sources: List[Source], query: str, limit: int) -> str:
|
|
"""Search files using Pinecone vector database."""
|
|
|
|
# Extract unique source_ids
|
|
# TODO: Inefficient
|
|
source_ids = [source.id for source in attached_sources]
|
|
if not source_ids:
|
|
return "No valid source IDs found for attached files"
|
|
|
|
# Get all attached files for this agent
|
|
file_agents = await self.files_agents_manager.list_files_for_agent(
|
|
agent_id=agent_state.id, per_file_view_window_char_limit=agent_state.per_file_view_window_char_limit, actor=self.actor
|
|
)
|
|
if not file_agents:
|
|
return "No files are currently attached to search"
|
|
|
|
results = []
|
|
total_hits = 0
|
|
files_with_matches = {}
|
|
|
|
try:
|
|
filter = {"source_id": {"$in": source_ids}}
|
|
search_results = await search_pinecone_index(query, limit, filter, self.actor)
|
|
|
|
# Process search results
|
|
if "result" in search_results and "hits" in search_results["result"]:
|
|
for hit in search_results["result"]["hits"]:
|
|
if total_hits >= limit:
|
|
break
|
|
|
|
total_hits += 1
|
|
|
|
# Extract hit information
|
|
hit_id = hit.get("_id", "unknown")
|
|
score = hit.get("_score", 0.0)
|
|
fields = hit.get("fields", {})
|
|
text = fields.get(PINECONE_TEXT_FIELD_NAME, "")
|
|
file_id = fields.get("file_id", "")
|
|
|
|
# Find corresponding file name
|
|
file_name = "Unknown File"
|
|
for fa in file_agents:
|
|
if fa.file_id == file_id:
|
|
file_name = fa.file_name
|
|
break
|
|
|
|
# Group by file name
|
|
if file_name not in files_with_matches:
|
|
files_with_matches[file_name] = []
|
|
files_with_matches[file_name].append({"text": text, "score": score, "hit_id": hit_id})
|
|
|
|
except Exception as e:
|
|
self.logger.error(f"Pinecone search failed: {str(e)}")
|
|
raise e
|
|
|
|
if not files_with_matches:
|
|
return f"No semantic matches found in Pinecone for query: '{query}'"
|
|
|
|
# Format results
|
|
passage_num = 0
|
|
for file_name, matches in files_with_matches.items():
|
|
for match in matches:
|
|
passage_num += 1
|
|
|
|
# Format each passage with terminal-style header
|
|
score_display = f"(score: {match['score']:.3f})"
|
|
passage_header = f"\n=== {file_name} (passage #{passage_num}) {score_display} ==="
|
|
|
|
# Format the passage text
|
|
passage_text = match["text"].strip()
|
|
lines = passage_text.splitlines()
|
|
formatted_lines = []
|
|
for line in lines[:20]: # Limit to first 20 lines per passage
|
|
formatted_lines.append(f" {line}")
|
|
|
|
if len(lines) > 20:
|
|
formatted_lines.append(f" ... [truncated {len(lines) - 20} more lines]")
|
|
|
|
passage_content = "\n".join(formatted_lines)
|
|
results.append(f"{passage_header}\n{passage_content}")
|
|
|
|
# Mark access for files that had matches
|
|
if files_with_matches:
|
|
matched_file_names = [name for name in files_with_matches.keys() if name != "Unknown File"]
|
|
if matched_file_names:
|
|
await self.files_agents_manager.mark_access_bulk(agent_id=agent_state.id, file_names=matched_file_names, actor=self.actor)
|
|
|
|
# Create summary header
|
|
file_count = len(files_with_matches)
|
|
summary = f"Found {total_hits} Pinecone matches in {file_count} file{'s' if file_count != 1 else ''} for query: '{query}'"
|
|
|
|
# Combine all results
|
|
formatted_results = [summary, "=" * len(summary)] + results
|
|
|
|
self.logger.info(f"Pinecone search completed: {total_hits} matches across {file_count} files")
|
|
return "\n".join(formatted_results)
|
|
|
|
async def _search_files_native(self, agent_state: AgentState, query: str, limit: int) -> str:
|
|
"""Traditional search using existing passage manager."""
|
|
# Get semantic search results
|
|
passages = await self.agent_manager.query_source_passages_async(
|
|
actor=self.actor,
|
|
agent_id=agent_state.id,
|
|
query_text=query,
|
|
embed_query=True,
|
|
embedding_config=agent_state.embedding_config,
|
|
)
|
|
|
|
if not passages:
|
|
return f"No semantic matches found for query: '{query}'"
|
|
|
|
# Limit results
|
|
passages = passages[:limit]
|
|
|
|
# Group passages by file for better organization
|
|
files_with_passages = {}
|
|
for p in passages:
|
|
file_name = p.file_name if p.file_name else "Unknown File"
|
|
if file_name not in files_with_passages:
|
|
files_with_passages[file_name] = []
|
|
files_with_passages[file_name].append(p)
|
|
|
|
results = []
|
|
total_passages = 0
|
|
|
|
for file_name, file_passages in files_with_passages.items():
|
|
for passage in file_passages:
|
|
total_passages += 1
|
|
|
|
# Format each passage with terminal-style header
|
|
passage_header = f"\n=== {file_name} (passage #{total_passages}) ==="
|
|
|
|
# Format the passage text with some basic formatting
|
|
passage_text = passage.text.strip()
|
|
|
|
# Format the passage text without line numbers
|
|
lines = passage_text.splitlines()
|
|
formatted_lines = []
|
|
for line in lines[:20]: # Limit to first 20 lines per passage
|
|
formatted_lines.append(f" {line}")
|
|
|
|
if len(lines) > 20:
|
|
formatted_lines.append(f" ... [truncated {len(lines) - 20} more lines]")
|
|
|
|
passage_content = "\n".join(formatted_lines)
|
|
results.append(f"{passage_header}\n{passage_content}")
|
|
|
|
# Mark access for files that had matches
|
|
if files_with_passages:
|
|
matched_file_names = [name for name in files_with_passages.keys() if name != "Unknown File"]
|
|
if matched_file_names:
|
|
await self.files_agents_manager.mark_access_bulk(agent_id=agent_state.id, file_names=matched_file_names, actor=self.actor)
|
|
|
|
# Create summary header
|
|
file_count = len(files_with_passages)
|
|
summary = f"Found {total_passages} semantic matches in {file_count} file{'s' if file_count != 1 else ''} for query: '{query}'"
|
|
|
|
# Combine all results
|
|
formatted_results = [summary, "=" * len(summary)] + results
|
|
|
|
self.logger.info(f"Semantic search completed: {total_passages} matches across {file_count} files")
|
|
|
|
return "\n".join(formatted_results)
|