Metadata-Version: 2.5
Name: noscium
Version: 0.0.2
Summary: Institutional memory for enterprise AI assistants — governed, multi-user, organizational memory framework.
Project-URL: Homepage, https://github.com/RageshRavi/noscium
Project-URL: Repository, https://github.com/RageshRavi/noscium
Project-URL: Issues, https://github.com/RageshRavi/noscium/issues
Author: Ragesh
License: MIT
License-File: LICENSE
Keywords: agents,ai,enterprise-ai,institutional-memory,llm,memory,organizational-memory,rag
Classifier: Development Status :: 2 - Pre-Alpha
Classifier: Intended Audience :: Developers
Classifier: License :: OSI Approved :: MIT License
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.9
Classifier: Programming Language :: Python :: 3.10
Classifier: Programming Language :: Python :: 3.11
Classifier: Programming Language :: Python :: 3.12
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
Classifier: Topic :: Software Development :: Libraries :: Python Modules
Requires-Python: >=3.9
Requires-Dist: anthropic>=0.40.0
Provides-Extra: dev
Requires-Dist: pytest>=8.0.0; extra == 'dev'
Provides-Extra: local-embeddings
Requires-Dist: sentence-transformers>=3.0.0; extra == 'local-embeddings'
Description-Content-Type: text/markdown

# Noscium

Institutional memory for enterprise AI assistants.

Unlike retrieval systems (RAG) or single-user memory libraries, Noscium is
designed for organizations: multiple users, teams, roles, and the governance
that comes with it.

> **Status: pre-alpha.** Phase 1 core (memory model, LLM-powered extraction,
> SQLite storage with scoped access control, relevance-ranked retrieval) has
> an initial working implementation with passing unit tests on the storage
> layer. Not yet hardened for production use — no consolidation, conflict
> resolution, or governance features yet (Phase 2/3).

## Why Noscium

Most AI agent memory systems optimize for *individual recall* — one user's
memories, retrieved well over time. Noscium's focus is different:
*organizational* memory — many users, teams, and roles, and the access
control, consolidation, and conflict resolution that come with that.

## Planned Features

- **Multi-user access control** — memories scoped by role and team
- **Organizational consolidation** — individual facts become org-level knowledge
- **Conflict resolution** — when memories contradict, track and resolve
- **Compliance & retention** — configurable policies, audit trails, provenance
- **Importance judging** — LLM + user feedback decides what's worth remembering
- **Memory types** — episodic (what happened), semantic (what is true), procedural (how to do X)

## Roadmap

| Phase | Focus | Status |
|---|---|---|
| Phase 1 | Core memory model, extraction, scoped retrieval | In progress |
| Phase 2 | Organizational consolidation, conflict resolution, memory graph | Planned |
| Phase 3 | Governance, compliance, retention policies, full audit trail | Planned |

## Installation

```bash
pip install noscium[local-embeddings]
```

(Pre-alpha — API will change without notice until a 0.1.0 release.)

## Quickstart

```python
import anthropic
from noscium import Noscium, VisibilityScope

client = Noscium(anthropic_client=anthropic.Anthropic())

# Alice tells the assistant something team-relevant
client.remember(
    "The Acme account prefers async standups over live meetings.",
    user_id="alice",
    team_id="support-team",
    visibility_scope=VisibilityScope.TEAM,
)

# Bob, on the same team, asks a related question later — sees Alice's memory
results = client.recall(
    "How does Acme like to communicate?",
    user_id="bob",
    team_id="support-team",
)
```

See `examples/quickstart.py` for a runnable version. Requires an
`ANTHROPIC_API_KEY` environment variable for the extraction step.

## Running Tests

```bash
pip install -e ".[dev]"
pytest tests/
```

The storage-layer tests run without any API key — they verify the
personal/team/org visibility scoping directly.

## License

MIT
