knowledge
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| KNOWLEDGE_DB | No | Path to the SQLite index database, relative to the config file. Defaults to index.db. | |
| KNOWLEDGE_BATCH | No | Embedding batch size. Defaults to 32. | |
| KNOWLEDGE_VAULT | No | Path to the Markdown vault. Overrides vault.path; required when no config file supplies it. | |
| KNOWLEDGE_CONFIG | No | Path to the knowledge.toml config file. Looked up in this order: --config, KNOWLEDGE_CONFIG, the current directory, then ~/.config/knowledge/. | |
| KNOWLEDGE_API_KEY | No | API key for the embeddings API. Only needed for semantic search; keyword search works without it. | |
| KNOWLEDGE_TIMEOUT | No | Embeddings request timeout in seconds. Defaults to 300. | |
| KNOWLEDGE_BASE_URL | No | Base URL of the OpenAI-compatible embeddings API. Defaults to https://api.openai.com/v1. | |
| KNOWLEDGE_EMBED_DIM | No | Embedding dimension. Defaults to 1536. | |
| KNOWLEDGE_TIER_DIRS | No | Tier 1 directories. Defaults to Memory and Notes. | |
| KNOWLEDGE_TIER_NAMES | No | Tier 1 file names. Defaults to CONTEXT.md, CONTEXT-*.md, MEMORY.md, INDEX.md. | |
| KNOWLEDGE_CREDENTIALS | No | Fallback credentials file containing KEY=value or KEY: value lines. Defaults to ~/.secrets.yaml. | |
| KNOWLEDGE_EMBED_MODEL | No | Embeddings model name. Defaults to text-embedding-3-small. |
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
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": false
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| knowledge_searchA | Гибридный поиск по базе знаний (ключевые слова + смысл). Возвращает файл, диапазон строк и фрагмент. Начинай с этого, а не с чтения файлов целиком. |
| knowledge_readA | Прочитать .md из базы целиком или диапазоном строк. Возвращает хеш — передай его в knowledge_write как expected_hash, чтобы не затереть чужие правки. |
| project_contextB | Стартовый контекст проекта: CONTEXT.md, MEMORY.md и файлы памяти плюс список чатов. Самый дешёвый способ войти в проект. |
| knowledge_writeA | Создать или перезаписать .md в базе. Если передать expected_hash из knowledge_read и файл изменился — вернётся conflict и записи не будет. expected_hash="new" требует, чтобы файла не было. |
| knowledge_appendB | Дописать текст в конец .md. Существующее содержимое не трогается. |
| knowledge_commitB | Сохранить заметку о сессии/решении. Кладёт новый файл в Projects/<проект>/Notes/ или в Notes/ с датой в имени. |
| knowledge_statusB | Состояние индекса: файлы, чанки, эмбеддинги, свежесть. |
| knowledge_projectsA | Список проектов в базе с числом файлов и чанков. |
| knowledge_reindexB | Пересобрать индекс. full=true игнорирует кеш и перечитывает все файлы. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 9 tools
Each tool has a distinct purpose: search, read, write, append, index status, project listing, reindexing, and a bootstrap context helper. The only mild overlap is between knowledge_commit (which creates a dated note file in Notes/) and knowledge_write (which creates/overwrites any .md), and between project_context and knowledge_read of the same bootstrap files, but the descriptions clearly mark commit as a convenience wrapper and project_context as the cheap entry point.
Eight of nine tools use a consistent knowledge_<verb> pattern (search, read, write, append, commit, reindex) with noun-style accessors for status/projects. The outlier is project_context, which drops the prefix and uses a noun_noun form, a minor but visible deviation.
Nine tools is well-scoped for a markdown knowledge base with search, indexing, and CRUD-style file access. Every tool maps to a distinct operation a caller would realistically need, with no filler.
The surface covers entry (project_context), discovery (search, projects), reading, writing, appending, and index maintenance (status, reindex), which is coherent lifecycle coverage. The notable gap is that there is no delete, rename, or move operation, so files can be created and overwritten but never removed or reorganized.