RAG
1760AI tools in the RAG category
efortune
FileTree-backed storage and retrieval substrate for agent memory and knowledge
rayadesu
Session-scoped context window switcher as a DSH composer control (maintained continuation of dsh-context-window-picker)
gustavogutierrez
Local semantic search engine for Alisio (LanceDB + multilingual-e5-small on ONNX) plus a Notebook app that answers only from your sources, with citations.
...moreskydeckai
An MCP Server for RAG Web Browser Actor
GitHub Actions
OSVA Connector Profile V1 Microsoft OneDrive MCP connector (delegated Graph v1 reads, bounded text retrieval, optional small upload when host-approved)
...morehfshfg
嵌入流水线的账与闸(不调模型、不联网):plan 按单批 token 上限分批并估算 token/成本/缓存命中,dedupe 两档去重(完全相同 sha256 + 近似 Jaccard,近似只给建议),cache 只存「内容 sha256 → 向量」不存原文,check 校验同库同维度与 L2 归一化、NaN 直接拒。token 估算模型显式(英 3.5 字/token、中 1.6 字/token 加权)并给误差带,单价读不到标 unknown 不填 0。零运行时依赖,含 CLI。 | Batch-em
...moreUntangle retrieval core: hybrid search (dense + deterministic BM25 + exact symbol names) with reciprocal rank fusion, explainable score breakdowns, complexity/findability filters, grep mode and deterministic output formats (spec.md §13, vision.md §1b/§1d)
...morechaoyangnz
`@harngine/retrieval` exposes a retriever as a recorded read Tool. The retriever chooses its index, ranking method, and storage backend.
...moreGitHub Actions
Mem0 long-term memory for DeepSeek Harness: managed Platform, self-hosted server or a local SQLite store (no Python), with a visual settings page, automatic recall/capture, editable memory summary, token budgets and weekly memory hygiene.
...morejalotis
Hinny: an embedded hybrid search database (full-text, vector and sparse retrieval with filters, in your process). Placeholder for the first public release.
...morevu.minh.quang
RAGO native MCP and local-first retrieval for Windows, macOS and Linux
GitHub Actions
Pi extension tools for agent self-reflection: model metadata, context window usage, thinking level, and reasoning budget.
...morehuakirax
Smart model configuration for DeepSeek Harness: add a provider baseURL + key, pull the upstream model list, auto-match wire protocol / context window / multimodal capabilities / reasoning effort from the pi-ai catalog and models.dev, and silently write it
...morechhetrisushil
Knowledge-graph MemoryStore decorator for Loom — graph-expanded retrieval plus a traversal API, over any MemoryStore, with learned edges persisted in the store you already have.
...moreezzabuzaid
A rule of an elt computed dataset: it cuts the text of each loaded row into chunks that each count at most a number of tokens of the embedding model's tokenizer, so each chunk can get its own vector.
...morepwie
Continuum: one memory plugin for DeepSeek Harness. A single SQLite store holding every source verbatim with provenance, three-way retrieval (lexical + bge-small vectors + activation decay), one prompt injection, one capture path for messages and tool trac
...moreGitHub Actions
A mutable, browser-safe retrieval index for documents an app owns: heading-aware chunks and SQLite FTS5 over any synchronous SQLite handle (node:sqlite, or sqlite-wasm on origin-private storage).
...moreGitHub Actions
Adaptive context compaction for DeepSeek Harness: derives each model's pressure threshold and retained tail from its context window and the session's measured growth.
...more