Platform and model notes
Verified 2026-10-12 against published wheels and the Conversational 0.7.19 release source. This page is not a measured latency report.
Install platforms
Section titled “Install platforms”Published Conversational, Coder, and Governance wheels cover:
- Windows x64
- Linux x86_64 with glibc 2.28 or newer
- macOS 15 or newer on Apple silicon
Python 3.10 or newer is required. Python 3.12 is a sensible starting point for a new environment.
Not listed as available: Intel Mac, Windows ARM, Linux ARM, Alpine, and Snap.
Structured uses the GitHub 0.7.3 wheel on Windows, PyPI 0.7.2 on Linux, and the GitHub 0.7.2 wheel on macOS 15 Apple silicon.
Package download size
Section titled “Package download size”Conversational 0.7.19 wheel sizes from the public PyPI JSON (binary MiB):
- Windows x64: 97 MiB
- Linux x86_64: 148 MiB
- macOS 15 Apple silicon: 96 MiB
Those numbers are the product package, not a model download. Conversational does not ship an answering model.
The model you connect
Section titled “The model you connect”You use your own AI. MemStrata keeps the index on your computer, selects passages, and shows them with the answer. The memory, graphics, and disk needs of that model are the model provider’s requirements, not a number published here.
Visual retrieval
Section titled “Visual retrieval”Optional ColQwen2 and ColFlor select original pages. They are not the answering model. EVIE and ColPali are coming soon and are not listed as supported downloads. Detecting a GPU is separate from installing a compatible runtime.
Evidence modes in the product
Section titled “Evidence modes in the product”Evidence tokens are the retrieved source budget. They are not the model’s full context window. They do not include system instructions, the question, later conversation, or output capacity.
| Mode | Evidence limit | When to use it |
|---|---|---|
| Standard | 8,192 tokens | Default for most questions |
| Deep | 24,000 tokens | When Standard evidence is not enough |
The R454 and R230 research runs used a 7,000-token evidence budget. That is a research protocol, not the product Standard or Deep setting.
Suggested starting path
Section titled “Suggested starting path”- Install Conversational 0.7.19 on a listed platform.
- Activate.
- Connect the AI you already use.
- Import the demo notes and follow the Conversational setup sequence.