librechat-mnemonic
by dowlings
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| HOST | No | Listen address | 0.0.0.0 |
| PORT | No | Listen port | 8710 |
| MCP_PATH | No | Where to serve the MCP endpoint | /mcp |
| LOG_LEVEL | No | pino level | info |
| UPSTREAMS | Yes | JSON array of upstream provider configs | |
| OLLAMA_URL | No | URL of the Ollama instance for embeddings | |
| DISABLE_GIT | No | Disable git in mnemonic | |
| EMBED_MODEL | No | Embedding model name | |
| MCP_ENABLED | No | Serve the MCP endpoint | true |
| MNEMONIC_TAG | No | Tag added to everything this service writes | librechat |
| MNEMONIC_URL | No | Required when MNEMONIC_MODE=remote | |
| EXTRACT_MODEL | No | Model name for extraction | |
| MNEMONIC_MODE | No | spawn runs the bundled mnemonic over stdio; remote connects to a streamable-http instance | spawn |
| EMBED_PROVIDER | No | Embedding provider for mnemonic | |
| GEMINI_API_KEY | No | Gemini API key for embeddings | |
| OPENAI_API_KEY | No | OpenAI API key for embeddings | |
| EXTRACT_API_KEY | No | Bearer token for extraction model | |
| EXTRACT_BASE_URL | No | OpenAI-compatible base URL, including /v1 | |
| MNEMONIC_COMMAND | No | Executable used in spawn mode | |
| MNEMONIC_HEADERS | No | JSON headers for the remote instance, e.g. auth | {} |
| MEMORY_WRITE_MODE | No | llm extracts automatically, explicit only on 'remember that …', off disables writing | llm |
| EXTRACT_TIMEOUT_MS | No | Extraction is detached; a timeout drops the write, never the reply | 30000 |
| LIBRECHAT_MONGO_DB | No | Override the database name | |
| MEMORY_PROJECTLESS | No | off disables memory entirely in chats not assigned to a project | global |
| LIBRECHAT_MONGO_URI | Yes | Connection string for LibreChat's MongoDB | |
| MEMORY_MAX_PER_TURN | No | Cap on memories written per exchange | 3 |
| MNEMONIC_TIMEOUT_MS | No | Per-call timeout | 20000 |
| MNEMONIC_VAULT_PATH | No | Vault directory, passed through as VAULT_PATH | /vault |
| MNEMONIC_WRITE_SCOPE | No | global stores in the main vault stamped with the project; project writes a project vault | global |
| LIBRECHAT_USER_HEADER | No | Header carrying {{LIBRECHAT_USER_ID}} | x-librechat-user-id |
| MEMORY_COMMAND_PREFIX | No | Change if it clashes with something | /memory |
| MEMORY_RECALL_ENABLED | No | Set false to write memories without injecting them | true |
| MNEMONIC_PROJECT_ROOT | No | Where per-project directories live | /projects |
| MNEMONIC_RECALL_LIMIT | No | Memories retrieved per turn | 6 |
| MNEMONIC_RECALL_SCOPE | No | project isolates, all boosts the current project, global returns the whole main vault | all |
| MEMORY_DEFAULT_ENABLED | No | Whether memory is on for chats with no explicit setting | true |
| MEMORY_DEDUPE_THRESHOLD | No | Recall score above which a candidate is treated as already known | 0.82 |
| MNEMONIC_MIN_SIMILARITY | No | Similarity floor passed to recall | 0.3 |
| MEMORY_MAX_CONTEXT_CHARS | No | Budget for the injected block | 4000 |
| MEMORY_QUERY_MESSAGE_COUNT | No | User turns used to build the recall query | 3 |
| LIBRECHAT_CONVERSATION_HEADER | No | Header carrying {{LIBRECHAT_BODY_CONVERSATIONID}} | x-librechat-conversation-id |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Server capabilities have not been inspected yet.
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
No tools | |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |