Skip to content

Roadmap capsule — SemeAI Gate Basic open-source path #2

Description

@SemeAIPletinnya

Purpose

This issue is the standing roadmap capsule for semeai-gate-basic.

It tracks the public/basic open-source path and prevents the repo from drifting into the full private SemeAI Local product.

Product wedge

Core wedge:

Stop unsupported AI answers before they reach users.

This repo should prove the smallest useful public version of SemeAI Gate:

User Message + AI Answer + Business Data + Business Rules + Business Context -> Gate -> SHOW / REVIEW / BLOCK

Internal mapping remains:

  • SHOW = PROCEED
  • REVIEW = NEEDS_REVIEW
  • BLOCK = SILENCE

Phase 0 — Clean public baseline

Goal: make the repository safe, portable, and reviewable.

Tasks:

  • Verify sanitized file list.
  • Confirm no private workspace data is included.
  • Confirm no private archives or receipt corpus are included.
  • Confirm README explains the product in simple language.
  • Confirm LICENSE exists.
  • Confirm package metadata exists.
  • Confirm examples run locally.
  • Confirm tests pass from a clean checkout.
  • Confirm schema version is documented.

Target result:

A clean public repo that can be shown to developers, design partners, and early B2B reviewers.

Phase 1 — Stable gate contract

Goal: freeze the minimal business-readable contract.

Tasks:

  • Version schema as 0.1.
  • Document all input fields.
  • Document all output fields.
  • Preserve SHOW / REVIEW / BLOCK mapping.
  • Add schema validation tests.
  • Add mapping tests.
  • Keep business-facing names separate from internal SaC/PoR names.

Required input fields:

  • user_message
  • ai_answer
  • business_data
  • business_rules
  • business_context
  • hidden_context_marker
  • business_risk
  • metadata

Required output fields:

  • action
  • internal_decision
  • show_to_user
  • reason
  • business_risk
  • context_integrity
  • risk_details
  • next_step
  • audit_id
  • audit_preserved
  • schema_version

Phase 2 — Deterministic examples

Goal: make the problem obvious without private data.

Required examples:

  • Fake promo code.
  • Unsupported financial or product claim.
  • Unsafe action.
  • Context drift.

Each example should show:

business pain -> AI answer -> gate action -> next step -> audit record

Phase 3 — SDK readiness

Goal: prepare for developer integration.

Tasks:

  • Python package API.
  • Optional Node wrapper or SDK surface.
  • JSON schema validation helper.
  • Local hash-only audit records.
  • Simple benchmark runner.
  • Copy-paste integration docs.

Phase 4 — Benchmark corpus

Goal: measure release-control behavior.

Tasks:

  • Deterministic labeled cases.
  • Expected actions.
  • False release tracking.
  • False block tracking.
  • Summary report.
  • Regression command for CI/local validation.

Phase 5 — Public demo readiness

Goal: make the repo understandable in under five minutes.

Tasks:

  • Quickstart.
  • Copy-paste integration example.
  • Simple architecture diagram.
  • Limitations section.
  • What this is not section.
  • Clean first public post/demo text.

What this repo is not

This repo is not:

  • a foundation model;
  • a chatbot UI;
  • the full SemeAI Local desktop app;
  • a universal hallucination detector;
  • a certified compliance system;
  • a replacement for human review;
  • a production enterprise platform by itself.

Current priority

Current working order:

  1. clean public baseline;
  2. stable schema;
  3. deterministic examples;
  4. tests;
  5. SDK/readme polish;
  6. GitHub-ready package;
  7. public demo only after the repo is clear and reproducible.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions