Status and development
Comm-SCI-Control as a JSON-only approach significantly improves adherence and transparency, but it cannot guarantee 100% deterministic compliance in plain chat environments. Therefore, a Python wrapper with API integration is being developed in parallel.
Core point: The JSON ruleset is a normative specification, not deterministic executable code.
Consequence: In chat-only usage, behavior remains probabilistic and dependent on model/system/context.
Main factors: model architecture, effective context window, prompt length, provider routing, and runtime load.
Practice: Anchor, Audit, QC deltas, and context-pressure guard reduce drift, but do not replace external enforcement.
| Model class | Assessment | Typical limitation |
|---|---|---|
| Large chat LLMs with large context windows | recommended | drift risk still exists in very long sessions |
| Mid-size models with limited context | partially suitable | earlier lost-in-the-middle and weaker contract fidelity |
| Small/legacy models | limited | unstable command parsing and higher rule-violation rates |
Privacy: Prompts may include sensitive information. Minimize/anonymize personal data before API submission.
Costs: API pricing is token-based; large rulesets plus long sessions increase spend significantly.
Recommendation: use compact artifacts, reduce prompt load, set budget caps, and control exports intentionally.
Machine-readable reference: versions/versions.json.
| Status | Meaning |
|---|---|
stable | recommended production line; actively maintained. |
supported | still usable; limited compatibility-oriented support. |
deprecated | kept for migration/comparison; no active feature development. |
experimental | test status; breaking changes possible. |
Maturity: advanced and already usable, but not yet production-ready.
Open work:
1. Main-program modularization is not complete yet (though already advanced).
2. Documentation and scientific tooling modules are still incomplete.
3. Formula rendering is still missing.
4. User document upload is still missing.
5. UI can be further optimized.
6. Systematic application tests and example collections are being expanded.
Contributions: collaborators are welcome. Development is currently done as a spare-time project.
Example views from the current wrapper state (GUI + panel, SCI activation, deep-dive output, anchor/audit).
