File Management
Attach files to memories and search file content using vector search.
Available Tools
| Tool |
Purpose |
Required Params |
| attach_files_to_memory |
Attach files to a memory |
memory_id, user_id, files[] |
| get_memory_files |
Vector search in files |
user_id, search_query |
Attaching Files
Tool: attach_files_to_memory
attach_files_to_memory({
memory_id: "abc-123",
user_id: "user-456",
files: [
{ "filename": "architecture.md", "content": "..." },
{ "filename": "api-spec.yaml", "content": "..." }
]
})
Parameters
| Param |
Type |
Required |
Description |
| memory_id |
string |
Yes |
Target memory to attach to |
| user_id |
string |
Yes |
File owner identifier |
| files |
array |
Yes |
Array of file objects |
File Object Format
{
"filename": "document-name.ext",
"content": "file content as string"
}
Searching Files
Tool: get_memory_files
get_memory_files({
user_id: 123,
search_query: "authentication flow diagram",
file_name: "architecture.md",
keywords: ["auth", "login", "JWT"],
limit: 5
})
Parameters
| Param |
Type |
Required |
Description |
| user_id |
int |
Yes |
File owner identifier |
| search_query |
string |
Yes |
Semantic search within file content |
| file_name |
string |
No |
Filter by specific filename |
| user_file_name |
string |
No |
Filter by user-facing filename |
| keywords |
string[] |
No |
Additional keyword filters |
| limit |
int |
No |
Max results (default 5) |
Common Workflows
Build Document Knowledge Base
1. add_memories(messages about project) -> get memory_id
2. attach_files_to_memory(memory_id, user_id, [design_doc, api_spec])
3. get_memory_files(user_id, "key topic") -> verify searchable
Search Across All Documents
get_memory_files({
user_id: 123,
search_query: "error handling patterns",
limit: 10
})
Search in Specific Document
get_memory_files({
user_id: 123,
search_query: "rate limiting",
file_name: "api-guidelines.md"
})
Keyword + Semantic Search
get_memory_files({
user_id: 123,
search_query: "how to handle retries",
keywords: ["retry", "backoff", "resilience"]
})
Best Practices
- Always provide user_id — required for all file operations
- Verify memory exists before attaching —
get_memory(memory_id) first
- Use descriptive filenames — helps with filtering later
- Semantic queries work best — use natural language, not just keywords
- Combine keyword + semantic — keywords filter, semantic search ranks
- Chunk large documents — split into topic-specific memories for better precision
- Organize by memory — attach related files to the same memory for grouping
Common Errors
| Error |
Cause |
Fix |
| File not found |
Invalid memory_id |
Verify memory exists first |
| Empty results |
Wrong user_id |
Check user_id matches the one used in attach |
| No matches |
Too specific query |
Broaden search query or remove file_name filter |
| Missing user_id |
Omitted required param |
Always include user_id for file operations |
agents-store