sfera/docs/roadmap/COMPLETION_PLAN.md

SFERA Completion Plan

Дата аудита: 2026-05-16.

Этот план фиксирует путь к завершению SFERA как продукта. Он не подменяет IMPLEMENTATION_ORDER.md, а добавляет проверяемые критерии готовности и текущий статус.

Текущее состояние

SFERA уже содержит рабочий backend/API, Rust BSL parser, SIR, semantic kernel, projection/incremental контур, runtime/governance/knowledge/collaboration пакеты и Next.js IDE.

Текущая проверенная база:

• Python/API tests: 219 passed.
• Rust tests: 15 passed.
• Frontend typecheck: passed.
• Frontend production next build: passed.
• Editor smoke: passed.
• Editor runtime smoke: passed.
• Shared Docker host smoke: passed for API, web, PostgreSQL, Neo4j, Qdrant, Redis and object storage on test-docker.
• Local Git baseline: initialized on branch main, first commit 3b845c8.

Definition Of Done

Проект считается завершённым для production-ready milestone, когда выполнены все пункты:

1. Все фазы из IMPLEMENTATION_ORDER.md имеют deterministic implementation без placeholder/stub.
2. Все parser bugs покрыты golden fixtures.
3. SIR snapshot validates для BSL/XML metadata projects.
4. Projection and incremental update pass against in-memory and Neo4j-compatible contracts.
5. API exposes project, semantic, governance, knowledge, collaboration and IDE workflows.
6. Frontend builds production bundle and passes editor smoke/runtime checks.
7. Stand runbook is executable on local/NAS and shared Docker host.
8. A single verification command can reproduce core/backend/Rust/frontend checks.

Phase Plan

Phase 1 — Semantic Core

Status: mostly complete.

Completed:

• packages/sir.
• rust/crates/bsl-parser.
• Rust query-parser foundation.
• Rust semantic-engine foundation.
• packages/semantic-kernel.
• packages/projection-engine.
• packages/incremental-indexer.
• Golden fixtures for mixed RU/EN syntax, calls, queries, writes, object writes and recordset writes.
• Same-module routine resolution priority for duplicate routine names.
• SIR diagnostics for malformed BSL routine/query boundaries.
• Incremental diagnostic refresh for changed BSL fragments.
• Routine Экспорт / Export flags preserved in SIR node attributes.
• BSL control-flow fixture for calls/writes inside loops and try/except blocks.
• Form event handler links from XML metadata to BSL routines.

Remaining:

• Keep parser and XML fixtures expanding as new 1C export patterns are encountered.

Phase 2 — Object Runtime

Status: implemented with tests, needs broader fixtures.

Completed:

• XML metadata fixtures preserve register dimensions/resources as SIR attributes.
• Semantic versioning classifies object rename/move/update changes.
• Object impact includes form event handlers and their register writes.
• XML fixtures cover child-element forms, commands and role rights.
• Query intelligence reports register read/write dependencies.
• Transaction topology groups routines by written register target.
• Object impact can include scheduled-job integration endpoints from the owning module.

Remaining:

• Add more impact/review fixtures for object-level dependency edge cases as new patterns are found.

Phase 3 — Runtime Intelligence

Status: implemented with tests, needs integration depth.

Remaining:

• Expand transaction/query intelligence with register movement patterns.
• Connect UI semantics to editor panels and review findings.
• Add scheduled job and integration cross-links to impact reports.

Phase 4 — Knowledge System

Status: implemented with tests.

Remaining:

• Add import fixtures for larger BSP/vendor packs.
• Add coverage reports by metadata subtree.

Phase 5 — Collaboration

Status: implemented with tests.

Remaining:

• Enforce task/session binding for mutations outside Phase 1 placeholders.
• Add UI flow coverage for comments/ownership/activity.

Phase 6 — Governance/Operations

Status: implemented with tests.

Remaining:

• Verify full stand runbook on test-docker.
• Add backup/restore and migration smoke checks.

Phase 7 — 1C IDE / Authoring

Status: working Next.js IDE exists and passes smoke; advanced authoring remains product work.

Completed:

• Backend and typed frontend API client for symbol search, go to definition and find references.

Remaining:

• Wire editor actions for go to definition, find references and symbol search into IDE panels.
• Complete semantic autocomplete from SIR + metadata context.
• Complete guarded apply with preview, semantic diff, impact, review, RBAC/privacy and version record.
• Complete object/form/report designer flows.
• Add browser smoke for key authoring flows.

AI Layer

Status: intentionally gated until semantic truth is stronger.

Remaining:

• Add answer/generation policy enforcement to every AI-facing endpoint.
• Add prompt/context builders over SIR, metadata, knowledge, privacy and task/session.
• Add preview-only AI code generation and guarded apply workflow.

Execution Order From Here

1. Connect the local Git baseline to the canonical Gitea repository on git.cin.su after the exact repository path is available.
2. Keep all current green checks green with scripts/check_all.ps1.
3. Reduce API test/runtime latency so the full Python suite stays comfortably below CI timeout budgets.
4. Finish Phase 1 hardening fixtures and diagnostics.
5. Expand Phase 2/3 fixtures around metadata, impact, query and runtime topology.
6. Finish guarded apply and object-level authoring API.
7. Finish IDE authoring UI and smoke coverage.
8. Add AI pair-programmer only after guarded apply is deterministic.
9. Verify full stand on local/NAS and shared Docker host.

Verification

Use:

.\scripts\check_all.ps1

Expected checks:

• cargo test && cargo build -p bsl-parser
• uv run pytest
• npm run typecheck
• next build
• optional frontend smoke checks when API/frontend stand is running.