mnemosyne-rag 0.6.1__tar.gz

mnemosyne_rag-0.6.1/LICENSE

@@ -0,0 +1,190 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to the Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   Copyright 2024 Jonathan Freed
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

mnemosyne_rag-0.6.1/PKG-INFO

@@ -0,0 +1,310 @@
+Metadata-Version: 2.4
+Name: mnemosyne-rag
+Version: 0.6.1
+Summary: A local, teaching-first RAG pipeline (Ollama + LangChain + FAISS) that turns any model into an instant expert.
+Author: Jonathan Freed
+License: Apache-2.0
+Project-URL: Homepage, https://github.com/freed-dev-llc/mnemosyne
+Project-URL: Repository, https://github.com/freed-dev-llc/mnemosyne
+Keywords: rag,ollama,langchain,faiss,retrieval,embeddings,local-llm
+Classifier: Development Status :: 3 - Alpha
+Classifier: License :: OSI Approved :: Apache Software License
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Requires-Python: >=3.11
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: langchain-core<2,>=1
+Requires-Dist: langchain-community<0.5,>=0.4
+Requires-Dist: langchain-ollama<2,>=1
+Requires-Dist: langchain-openai<2,>=1
+Requires-Dist: langchain-text-splitters<2,>=1
+Requires-Dist: pypdf>=4.2
+Requires-Dist: pydantic-settings>=2.4
+Requires-Dist: pyyaml>=6.0
+Requires-Dist: typer>=0.12
+Requires-Dist: rich>=13.7
+Requires-Dist: mcp>=1.28.0
+Requires-Dist: fastapi>=0.110
+Requires-Dist: uvicorn>=0.29
+Provides-Extra: cpu
+Requires-Dist: faiss-cpu>=1.8; extra == "cpu"
+Provides-Extra: dev
+Requires-Dist: ruff>=0.6; extra == "dev"
+Requires-Dist: pytest>=8.2; extra == "dev"
+Requires-Dist: pytest-cov>=5; extra == "dev"
+Requires-Dist: mypy>=1.11; extra == "dev"
+Requires-Dist: types-PyYAML>=6.0; extra == "dev"
+Requires-Dist: httpx>=0.27; extra == "dev"
+Dynamic: license-file
+
+# Mnemosyne
+
+![Mnemosyne Recall](docs/assets/mnemosyne/logos/mnemosyne_logo_horizontal.svg)
+
+[![CI](https://github.com/freed-dev-llc/mnemosyne/actions/workflows/ci.yml/badge.svg)](https://github.com/freed-dev-llc/mnemosyne/actions/workflows/ci.yml)
+[![codecov](https://codecov.io/gh/freed-dev-llc/mnemosyne/graph/badge.svg)](https://codecov.io/gh/freed-dev-llc/mnemosyne)
+[![Python 3.11+](https://img.shields.io/badge/python-3.11%2B-blue.svg)](https://www.python.org/downloads/)
+[![License: Apache-2.0](https://img.shields.io/badge/license-Apache--2.0-blue.svg)](LICENSE)
+[![Linting: ruff](https://img.shields.io/badge/linting-ruff-261230.svg)](https://github.com/astral-sh/ruff)
+[![Stack: Ollama · LangChain · FAISS](https://img.shields.io/badge/stack-Ollama%20·%20LangChain%20·%20FAISS-A78BD0.svg)](#3-technology)
+
+> **Mnemosyne** (*nee-MAH-suh-nee*): Titaness of memory, mother of the nine Muses. A
+> local Retrieval-Augmented Generation (RAG) pipeline that turns any model into an instant
+> expert on documents it has never seen, no fine-tuning required.
+
+Mnemosyne is a **teaching-first RAG pipeline** built on
+**[Ollama](https://ollama.com) + [LangChain](https://www.langchain.com) +
+[FAISS](https://github.com/facebookresearch/faiss)**: entirely local, no API keys, no
+data leaving the box. Point it at a corpus and it embeds and indexes that corpus; a small
+local model then answers questions about it *with citations*, as if it had read every
+page. In Greek myth, Mnemosyne was memory itself, the wellspring the Muses drew knowledge
+from. That's the job here: be the memory a model retrieves from.
+
+**Part of the freed-dev-llc family.** Mnemosyne is the **knowledge brain**: its first real use
+case is being a **local expert for [Argus](https://github.com/freed-dev-llc/argus)** (a network
+source-of-truth), fed by the same vendors Argus discovers, starting with **Ubiquiti**.
+
+[![Argus: freed-dev-llc/argus](docs/assets/buttons/btn_argus.svg)](https://github.com/freed-dev-llc/argus)
+
+---
+
+## Start Here
+
+| Need | Read | Tokens (~) |
+|------|------|------------|
+| **What RAG is, from first principles** | [`docs/RAG-101.md`](docs/RAG-101.md): the teaching core (chunk → embed → index → retrieve → generate) | ~4k |
+| **How the pipeline is wired** | [`docs/ARCHITECTURE.md`](docs/ARCHITECTURE.md): modules, data flow, where to change things | ~3k |
+| **Knowledge packs (the plugin model)** | [`docs/KNOWLEDGE_PACKS.md`](docs/KNOWLEDGE_PACKS.md): how a corpus becomes an expert; build your own | ~2k |
+| **What's planned** | [`docs/ROADMAP.md`](docs/ROADMAP.md): Ubiquiti corpus maturity, then vendor-pack parity with Argus | ~1k |
+| **Design decisions** | [`docs/architecture/adr/`](docs/architecture/adr/): why Ollama/LangChain/FAISS, why packs | ~2k |
+
+---
+
+## Table of Contents
+
+1. [The idea](#1-the-idea)
+2. [How it works](#2-how-it-works)
+3. [Technology](#3-technology)
+4. [Quickstart](#4-quickstart)
+5. [Knowledge packs](#5-knowledge-packs)
+6. [Use case: a local brain for Argus](#6-use-case-a-local-brain-for-argus)
+7. [Repository layout](#7-repository-layout)
+8. [Roadmap](#8-roadmap)
+9. [Contributing](#9-contributing)
+10. [License](#10-license)
+
+---
+
+## 1. The idea
+
+A base model knows a lot of *general* things and very few of *your* things. Fine-tuning to
+fix that is slow, expensive, and stale the moment your docs change. **Retrieval-Augmented
+Generation** takes the other road: leave the model's weights alone and, at question time,
+*retrieve* the few passages that actually matter and hand them to the model as context.
+The model doesn't need to have *memorized* the Ubiquiti switching guide; it just needs it
+in front of it for the next 4,000 tokens.
+
+Mnemosyne exists to make that loop **legible**. Every stage is a small, readable module
+you can open, change, and measure. It is a repo to *learn RAG by building it*, not a
+black box, and to grow into something useful for the family.
+
+> **Credit.** The pipeline shape and defaults follow Mariya Sha's excellent
+> [rag_ollama](https://github.com/MariyaSha/rag_ollama) tutorial
+> ([video](https://youtu.be/oZYlrooPgvs)): Mnemosyne reworks that notebook crash-course
+> into an installable, knowledge-pack-based package (no notebook), keeps its Ollama +
+> LangChain + FAISS stack and tiny-model defaults, and adds the pack framework, CLI, and
+> family scaffolding. Her demo corpus (a delightful "is Lord Elrond secretly Agent Smith?"
+> investigation) lives in her repo; Mnemosyne ships its own Ubiquiti example instead.
+
+## 2. How it works
+
+```
+                          INGEST (build the memory)
+  documents ─► load ─► chunk ─► embed (Ollama) ─► FAISS index ─► knowledge/<pack>/index.faiss
+  (md/pdf/html/txt)            (bge-m3)                          (saved to disk, reusable)
+
+                          ASK (retrieve + generate)
+  question ─► embed ─► FAISS top-k search ─► stuff context + question into prompt
+                                                   │
+                                                   ▼
+                              Ollama LLM (llama3.1 / qwen2.5 …) ─► answer + citations
+```
+
+Two paths, one index. **Ingest** is run once per corpus (and re-run when docs change);
+it is the slow, expensive part. **Ask** is cheap and local: embed the question, pull the
+nearest chunks out of FAISS, and let a small Ollama model write the answer grounded in
+exactly those chunks. Nothing leaves the machine. If nothing in the index is close enough
+to the question, Ask answers "not in the knowledge base" rather than reaching for unrelated
+chunks: a relevance floor (`MNEMOSYNE_SCORE_FLOOR`, on by default; set it to `none` to
+disable) keeps off-topic questions from getting confident, ungrounded answers.
+
+See [`docs/RAG-101.md`](docs/RAG-101.md) for the *why* behind every box, and
+[`docs/ARCHITECTURE.md`](docs/ARCHITECTURE.md) for which module owns each one.
+
+## 3. Technology
+
+| Layer | Choice | Why |
+| --- | --- | --- |
+| **LLM + embeddings** | [Ollama](https://ollama.com) | Local, free, swappable models; one runtime for chat (`qwen2.5:1.5b`) *and* `bge-m3` embeddings. |
+| **Orchestration** | [LangChain](https://www.langchain.com) | Batteries-included loaders, splitters, retrievers, and prompt plumbing: the teaching scaffold. |
+| **Vector store** | [FAISS](https://github.com/facebookresearch/faiss) | Fast, file-based, zero-service similarity search; the index is just a file you can ship. |
+| **Environment** | [mamba](https://github.com/mamba-org/mamba) / conda | Reproducible env that ships FAISS (CPU *and* GPU) from conda-forge; swap one file to go GPU. |
+| **Packaging** | Python 3.11+, `pyproject.toml`, `ruff` | Installable `mnemosyne` package + `mnemosyne` CLI; family lint/CI baseline. |
+
+The chat backend is also swappable: set `chat_provider=openai` (env `MNEMOSYNE_CHAT_PROVIDER`)
+and generation can target any OpenAI-compatible server, such as vLLM, llama.cpp's server, or
+LM Studio ([ADR-0009](docs/architecture/adr/0009-selectable-chat-backend.md)). Embeddings stay
+Ollama-only, and the default (`chat_provider=ollama`) is unchanged, so the Ollama row above
+stays accurate as written.
+
+> Defaults are tiny and CPU-friendly, following the [rag_ollama](https://github.com/MariyaSha/rag_ollama)
+> tutorial this project is based on: `bge-m3` for embeddings and `qwen2.5:1.5b` for
+> generation, with `chunk_size=500 / chunk_overlap=150 / k=5`. Override per-pack (in a
+> pack's `manifest.yaml`) or globally via `.env`; see [`.env.example`](.env.example).
+
+## 4. Quickstart
+
+**Prerequisites:** [Ollama](https://ollama.com/download) running locally, and `mamba`:
+the simplest way to get it is [Miniforge](https://github.com/conda-forge/miniforge), which
+bundles conda + mamba with conda-forge preconfigured (plain conda works too).
+
+```bash
+# 0. Pull the default models (once)
+ollama pull bge-m3          # embeddings
+ollama pull qwen2.5:1.5b    # tiny chat model
+
+# 1. Create the environment with mamba and install Mnemosyne
+mamba env create -f environment.yml      # GPU? use environment-gpu.yml instead
+mamba activate mnemosyne
+
+# 2. Build the memory for a knowledge pack (embeds + indexes its corpus)
+mnemosyne ingest ubiquiti
+
+# 3. Ask the expert
+mnemosyne ask ubiquiti "How do I adopt a UniFi switch to a remote controller?"
+
+# Or explore interactively (with chat history)
+mnemosyne chat ubiquiti
+```
+
+> **No conda?** A bare-pip path works too: `pip install -e ".[dev,cpu]"` (the `cpu` extra
+> pulls `faiss-cpu` from PyPI). The mamba path is preferred because it also unlocks GPU
+> FAISS by swapping in `environment-gpu.yml`; see [ADR-0004](docs/architecture/adr/0004-conda-mamba-environment.md).
+
+`ingest` writes a reusable FAISS index under `knowledge/`; `ask` and `chat` load it and
+answer with inline `[source]` citations. List what's available with `mnemosyne packs`.
+
+```bash
+mnemosyne packs            # list installed knowledge packs and index status
+mnemosyne ingest --help    # chunking / embedding-model overrides
+mnemosyne ask --help       # top-k, model, show-sources flags
+mnemosyne eval --help      # retrieval hit-rate + answer faithfulness against a labelled question set
+mnemosyne sweep --help     # chunk-size / k / model sweeps to pick defaults from data
+```
+
+### Use from a coding agent (MCP)
+
+Mnemosyne ships an MCP stdio server (`mnemosyne-mcp`) so MCP clients (Claude Code, and the
+agents behind Argus) can ground answers in its packs. A local
+`.mcp.json` registers it; the tools are:
+
+- `list_packs`: packs + whether each has a built index
+- `ask(pack, question, k?)`: a grounded answer with cited sources
+- `search(pack, query, k?)`: the top-k raw chunks, to reason over yourself
+
+The server honors `MNEMOSYNE_OLLAMA_HOST` from its environment. Because some MCP clients spawn
+servers with a stripped environment, set it explicitly when needed:
+
+```jsonc
+// .mcp.json
+{ "mcpServers": { "mnemosyne": { "type": "stdio", "command": "mnemosyne-mcp",
+    "env": { "MNEMOSYNE_OLLAMA_HOST": "http://localhost:11434" } } } }
+```
+
+### Use from a web UI or service (HTTP)
+
+Web UIs (e.g. an "ask the brain" box in the Argus dashboard) can't speak MCP, so Mnemosyne
+also ships a FastAPI server (`mnemosyne-http`, default `127.0.0.1:8088`):
+
+```bash
+mnemosyne-http      # GET /health · GET /packs · POST /ask · POST /search
+curl -s localhost:8088/ask -H 'content-type: application/json' \
+  -d '{"pack":"ubiquiti","question":"How do I adopt a switch remotely?"}'
+```
+
+`POST /ask` → `{answer, sources[]}`; `POST /search` → `{results[]}`. Intended to be called
+**server-to-server** (e.g. an Argus backend proxies to it), so there's no browser CORS to
+manage. Bind address is `MNEMOSYNE_HTTP_HOST` / `MNEMOSYNE_HTTP_PORT`. Interactive API docs are
+at `/docs`, and the service serves its app icon at `/favicon.svg`.
+
+To run it as a managed service (systemd) and let a remote consumer such as Argus reach it over a mesh network, see [`deploy/README.md`](deploy/README.md).
+
+## 5. Knowledge packs
+
+A **knowledge pack** is a self-contained expert: a corpus plus a small manifest describing
+how to load, chunk, and prompt it. The design deliberately mirrors **Argus vendor packs**
+([ADR-0005](https://github.com/freed-dev-llc/argus/blob/main/docs/architecture/adr/0005-vendor-packs.md)):
+each pack is discovered via a `mnemosyne.knowledge_packs` entry point, so packs can live
+in-tree or ship out-of-tree and independently.
+
+```
+src/mnemosyne/packs/ubiquiti/
+├── manifest.yaml     # name, models, chunking, sources, system prompt
+├── pack.py           # the KnowledgePack implementation (loaders + metadata)
+└── sources/          # URLs / local docs that make up the corpus
+```
+
+Two packs ship in-tree: **`ubiquiti`** (the technical worked example) and **`general`**, a
+curated corpus of operating principles and decision heuristics, the project-agnostic
+counterpart that grounds *how to decide* rather than *how a technology works*. To build your
+own, copy a pack directory, edit
+the manifest, point it at your docs, and `mnemosyne ingest <yourpack>`. See
+[`docs/KNOWLEDGE_PACKS.md`](docs/KNOWLEDGE_PACKS.md).
+
+## 6. Use case: a local brain for Argus
+
+[Argus](https://github.com/freed-dev-llc/argus) keeps a network's *truth* in NetBox by
+discovering devices through **vendor packs** (UniFi in-tree today). But discovery answers
+*"what is on the network."* It doesn't answer *"how does this technology actually work,
+and what should I do about it?"* That second question needs an **expert**, and that expert
+needs the vendor's documentation at its fingertips.
+
+Mnemosyne is that expert. The plan is to **mirror Argus's vendor packs as Mnemosyne
+knowledge packs**: Argus discovers Ubiquiti, Mnemosyne *explains* Ubiquiti, so an MCP
+agent can ask "Argus, what's on the network?" and "Mnemosyne, how do I fix this UniFi
+adoption loop?" and get grounded answers to both. Ubiquiti is the shared starting point;
+the two repos expand vendor coverage in step. See [`docs/ROADMAP.md`](docs/ROADMAP.md).
+
+## 7. Repository layout
+
+| Path | What |
+| --- | --- |
+| `src/mnemosyne/` | The pipeline: `loaders`, `chunking`, `embeddings`, `index`, `pipeline`, `prompts`, `cli` |
+| `src/mnemosyne/packs/` | Knowledge-pack framework (`base`, `registry`) + the in-tree `ubiquiti` pack |
+| `docs/` | `RAG-101` (teaching), `ARCHITECTURE`, `KNOWLEDGE_PACKS`, `ROADMAP`, and ADRs |
+| `knowledge/` | Built FAISS indices + raw corpora (gitignored; rebuilt by `ingest`) |
+| `examples/` | Minimal scripts showing the API end-to-end |
+| `tests/` | Pytest suite (chunking, index, pack discovery) |
+| `.github/` | CI, Dependabot, issue/PR templates, CODEOWNERS |
+
+## 8. Roadmap
+
+Mnemosyne started as a *teaching* pipeline and is growing toward a *useful* one. Full
+detail in [`docs/ROADMAP.md`](docs/ROADMAP.md):
+
+- **v0.1: the loop works (shipped).** `ingest` → `ask`/`chat` end-to-end on the Ubiquiti pack with citations.
+- **v0.2: measure it (shipped).** Eval harness (`mnemosyne eval`): retrieval hit-rate and answer faithfulness, with a CI regression gate.
+- **v0.3: serve it (shipped).** MCP (`mnemosyne-mcp`) and HTTP (`mnemosyne-http`) transports so coding agents (e.g. Argus) can call Mnemosyne as a tool.
+- **v0.x: vendor-pack parity with Argus (current).** Maturing the Ubiquiti corpus and expanding knowledge packs in lockstep with Argus discovery.
+
+## 9. Contributing
+
+This is a personal project run with the discipline of a shared one: changes land via
+**signed-commit pull requests** with a CHANGELOG entry, decisions are captured as
+**ADRs**, and work is tracked in **issues**. See [CONTRIBUTING.md](CONTRIBUTING.md).
+
+## 10. License
+
+[Apache-2.0](LICENSE).