Animus Roadmap¶

Project: Animus — Personal AI exocortex / Mind-class system Classification: Flagship Version: 2.3.0 (core), mixed across packages (see Build Truth) Last updated: 2026-06-29 Next review: 2026-07-27

Build Truth¶

⚠️ The previous roadmap overstated completion. This version is grounded in filesystem reality.

Package	Version	Source Files	Tests	Runtime Status
`core`	2.3.0	94 Python	2,832	CLI + memory + dashboard operational
`forge`	1.9.0	352 Python	10,431	Workflow orchestration operational
`bootstrap`	0.8.0	136 Python	1,874	Daemon + onboarding + dashboard operational
`kernel`	0.1.0	189 Python	107	Extracted from Forge, durable core wired
`quorum`	1.2.0 (convergentAI)	14 Rust + 10 Python	961	Rust core + Python bindings, active
`types`	0.1.0	4 Python	67	Minimal but functional
`pwa`	—	19 TypeScript	0	Frontend scaffolded, wiring TBD
`contracts`	—	20 JSON schemas	0	Pure JSON, no runtime validation
`database`	—	1 Python + 1 SQL migration	4	Migration syntax tests (no DB required)

Version alignment: ⚠️ Documented mismatch. Core 2.3.0, Forge 1.9.0, Bootstrap 0.8.0, others 0.1.0–1.2.0. See COMPATIBILITY_MATRIX.md for dependency graph and compatibility ranges. Unified versioning deferred to Phase 2.

Truth baseline: ✅ Honest. 25/26 PASS, 1 FAIL (version_alignment — expected and documented). Previously: 21/22 with 1 FAIL. Test infrastructure fixed: root pyproject.toml now sets pythonpath for all sibling packages.

Current State (Real)¶

What Works¶

Core exocortex: CLI (python -m animus), memory tiers (SQLite/ChromaDB), identity, proactive tasks
Bootstrap: Install daemon, onboarding wizard, FastAPI+HTMX dashboard (localhost:7700), Ollama health checks
Forge: Multi-agent YAML workflows, 10 archetypes, token budgets, checkpoint/resume, SQLite state, adversarial test harness, governance plane
Quorum: Rust intent graph + Python bindings, stigmergy coordination, signal bus, triumvirate voting
Kernel: Extracted 172 files / 55K LOC from Forge, budget/executor/sandbox/resume all present, durable core wired
PWA: TypeScript frontend scaffolded with Vite, login/chat/personas/status views, service worker
Types: Sensitivity, egress, secrets dataclasses (security-focused)
Contracts: 20 canonical JSON schemas (action, event, assessment, memory, trace, etc.)

What Is Scaffolding Only¶

Structure	Status	Evidence
`apps/`	✅ README only	`apps/README.md` explains what belongs here
`modules/`	✅ README only	`modules/README.md` explains boundary vs packages/
`contracts/` (root)	✅ Removed	Root dir deleted; `packages/contracts/` holds schemas
`database/`	✅ Migration + README	`database/README.md` + `migrations/versions/001_initial_schema.py`
`infra/`	✅ Docker Compose + README	`infra/docker-compose.yml` + `infra/.env.example`
`evidence/releases/`	✅ Bundles generated	`evidence-2026-06-29-075813/` with manifest, tests, schemas, git info
`adrs/`	✅ 5 ADRs	ADR-001 through ADR-005 all committed

Milestones (Revised)¶

Phase 0: Foundation — MOSTLY COMPLETE¶

Dead packages archived to packages/_archive/ (8 packages)
ADR-001 through ADR-005 recorded in adrs/
20 canonical JSON schemas extracted to packages/contracts/
Truth baseline fixed — check_test_count now FAILs on non-zero exit codes; python3 fallback added
Version alignment check implemented — reads all 8 package versions, reports mismatches (still 7 unique versions)
Schema importability — 20 schemas auto-generated as Pydantic v2 models in packages/types/
Root architecture dirs — populated with READMEs or removed (root contracts/ deleted)

Blockers: Phase 1 can start. Remaining Phase 0 gap: unify package versions (not urgent).

Phase 1: Schema Integration (Q3 2026)¶

Port 20 JSON schemas into importable Python types in packages/types/ (via scripts/compile_schemas.py)
Add pyproject.toml to packages/contracts/ (now a real package)
Build schema validation layer (JSON Schema → Pydantic or dataclass)
Add JSON Schema CI gate to docs-validate.py or new CI job
Document schema usage patterns in docs/reference/
Align packages/quorum/ name (convergentAI → animus-quorum or document exception)

Phase 2: Durable Core (Q3–Q4 2026) — COMPLETE¶

Implement object registry in database/ (PostgreSQL default, SQLite projection)
Implement event ledger with bitemporal projections
Wire packages/kernel/ into durable core model (DurableMemoryStore in memory/stores/durable.py)
Add traceability linter (scripts/traceability_linter.py)
Populate database/ with migration tooling
Populate infra/ with Docker Compose or systemd deployment manifests

Phase 3: Evidence & Governance (Q4 2026) — COMPLETE¶

Build scripts/assemble_evidence_bundle.py
Create release evidence bundles in evidence/releases/ (evidence-2026-06-29-075813)
Integrate adversarial test harness (packages/forge/src/animus_forge/evaluation/adversarial.py)
Implement policy decision point + governance plane (packages/forge/src/animus_forge/governance/)
Add ADR-002 through ADR-005 for decisions made to date

Phase 4: Public/Private Split (Q1 2027) — COMPLETE (Documentation + Decision)¶

Initialize private repo for owner-specific data — spec ready, migration guide written
Move synthetic fixtures to public repo — memory_eval_corpus.json already synthetic and public-safe
Document boundary and migration path — docs/operators/public-private-split.md + migration-guide.md
Make repo public OR deploy docs to Netlify/Cloudflare Pages — ADR-006 recommends public; Cloudflare Pages workflow provided as fallback

Phase 5: Mind Action Pack Integration (Q1–Q2 2027)¶

Port .claude/agents/, .claude/skills/, .claude/rules/ from artifacts into versioned packages
Validate against current operating model (v2.1 planes)
Document agent runtime expectations in docs/operators/

Prioritized Next Actions (Post-Phase 0)¶

Schema validation layer — JSON Schema → Pydantic gate in CI for packages/contracts/
Kernel durable core wiring — connect packages/kernel/ memory backends to database/ registry/ledger
Traceability linter — requirement-to-test mapping from ADRs to test modules
Evidence bundle release — generate first official bundle in evidence/releases/
Quorum naming alignment — document or rename convergentAI → animus-quorum

Blockers¶

Blocker	Impact	Resolution
Version misalignment	Low	Documented in `COMPATIBILITY_MATRIX.md`; unified versioning deferred
Kernel under-tested (99 tests for 189 files)	Medium	Add coverage before durable core integration
GitHub CI billing blocked	Medium	All CI jobs fail; local-first validation only
MkDocs deployment blocked	Low	Repo is private; Pages requires Pro or public repo
Alembic `sqlalchemy.url` configuration	Low	Default `driver://user:pass@...` in `alembic.ini` — set via env var or local config

Definition of Done (Phase 0 Completion)¶

Truth baseline fails on broken checks, not silently passes
All 8 packages have version metadata and alignment is documented (COMPATIBILITY_MATRIX.md)
packages/types/ imports at least 10 of the 20 contract schemas as Python
Every root architecture directory has a README.md explaining its purpose or is removed
scripts/assemble_evidence_bundle.py exists and produces a manifest

MkDocs Deployment¶

Status: Site builds clean (0 warnings). Deployment blocked because repo is private.
Options: Make repo public (fastest), upgrade GitHub Pro, or deploy to Netlify/Cloudflare Pages
Files: mkdocs.yml, docs/requirements-docs.txt, .github/workflows/docs-deploy.yml all ready