ADAAD — Autonomous Development & Adaptation Architecture: governance-first constitutional AI evolution runtime
Project description
⚡ Quickstart · 📜 The 23 Rules · 🗺 Roadmap · 🤖 Agents · 📱 Android · 🔬 Examples · 📊 Aponi · 📋 Changelog
ADAAD is a constitutionally governed code evolution engine — every mutation cryptographically signed, hash-chained into a tamper-evident ledger, deterministically replayable, and gated through 23 constitutional rules before a single byte changes.
It has already evolved itself. March 13, 2026. The hash is in the ledger.
Why ADAAD Is Different
Every AI code tool works the same way: suggest → you apply. No audit trail. No fitness score. No rollback. You are the last line of defense. ADAAD inverts this — the GovernanceGate is the last line of defense, not you.
|
|
|
|
ADAAD vs. Everything Else
⛓ Phase 65 — First Autonomous Self-Evolution
March 13, 2026. Zero human intervention in the execution path. Full human oversight of the constitutional framework. The hash is in the ledger.
View full evidence ledger entry (JSON)
{
"event": "EPOCH_COMPLETE",
"phase": 65,
"version": "9.0.0",
"date": "2026-03-13",
"constitutional_invariants_enforced": [
"MUTATION-TARGET", "AUDIT-0", "REPLAY-0", "CEL-ORDER-0",
"CEL-EVIDENCE-0", "SANDBOX-DIV-0", "GATE-V2-EXISTING-0"
],
"execution_trace": {
"01_TARGET_DISCOVERY": "✓ CodeIntelModel → highest-priority Tier-1 capability",
"02_CAPABILITY_MAP": "✓ CapabilityDiscovery → cap:runtime.evolution.cel_wiring.live_mode",
"03_PROPOSAL_GENERATE": "✓ ProposalEngine → AST mutation — proposal_hash recorded",
"04_STATIC_SCAN": "✓ StaticSafetyScanner → all 4 AST rules PASS",
"05_SANDBOX_TOURNAMENT": "✓ SandboxTournament → top candidate in ephemeral container",
"06_FITNESS_SCORE": "✓ FitnessEngineV2 → 7/7 signals above baseline",
"07_GOVERNANCE_GATE_V2": "✓ GovernanceGateV2 → APPROVED (Class A)",
"08_GOVERNANCE_GATE": "✓ GovernanceGate → CONFIRMED — all 23 rules pass",
"09_PATCH_APPLY": "✓ ASTDiffPatch → atomic apply — replay_verifier: 0 divergences",
"10_CAPABILITY_UPDATE": "✓ CapabilityGraph → target bumped, CapabilityChange to ledger",
"11_EPOCH_EVIDENCE": "✓ EpochEvidence → SHA-256 hash-chained into evolution ledger",
"12_HUMAN_OVERSIGHT": "✓ Full trail reviewed in Aponi — acknowledgement hash committed"
},
"proof": {
"replay_divergences": 0,
"governance_bypasses": 0,
"retroactive_evidence": false,
"silent_failures": 0
}
}
📅 75-Phase Evolution Timeline
| Phase | Era | Milestone | Version |
|---|---|---|---|
| 1–2 | Foundation | GovernanceGate · Evidence Ledger · SHA-256 hash-chaining |
v1.0 |
| 3–4 | Adaptive Intelligence | AdaptiveWeights EMA · SemanticDiffEngine AST analysis |
v2.x |
| 5–6 | Federation & Autonomy | Multi-repo federation · HMAC gating · roadmap self-amendment | v3.x |
| 7–8 | Governance Calibration | Reviewer reputation · governance health dashboard | v3.2–3.3 |
| 9–56 | Core Build-Out | 48 phases · Hardening · Evidence · Scale · Intelligence | v4.x–8.x |
| 57 | Keystone | ProposalEngine auto-provisioning | v8.0 |
| 58 | Perception | CodeIntelModel — code intelligence layer | v8.1 |
| 59 | Identity | CapabilityGraph v2 + CapabilityTargetDiscovery | v8.2 |
| 60 | Motor | AST Mutation Substrate + SandboxTournament | v8.3 |
| 61 | Evolution | Lineage Engine + CompatibilityGraph | v8.4 |
| 62 | Intelligence | MultiHorizon FitnessEngine v2 | v8.5 |
| 63 | Judgment | GovernanceGate v2 + Exception Tokens | v8.6 |
| 64 | Selfhood | Constitutional Evolution Loop (CEL) + EpochEvidence | v8.7 |
| 65 | Emergence ⛓ | First Autonomous Self-Evolution — March 13, 2026 | v9.0 |
| 66 | Hardening Tier Alpha | Telemetry completeness · lineage invariants · governance contracts | v9.1 |
| 67–68 | Operator Control | Operator-gated deployment · LLM failover governance | v9.2–9.3 |
| 69–70 | Innovations Router | Seed ingestion pipeline · innovation scoring | v9.4–9.5 |
| 71–72 | Seed Lifecycle | Seed promotion queue · graduated seed tracking | v9.6–9.7 |
| 73 | Seed Governance | Human-governed seed review · audit:write scope · Aponi review panel |
v9.8 |
| 74 | Seed → Proposal | Seed-to-Proposal Bridge · deterministic cycle_id routing |
v9.9 |
| 75 | Seed CEL Injection | Approved seeds wired into CEL Step 4 as advisory signal | v9.10.0 |
| 76 | Seed CEL Outcome Recorder | CEL epoch outcome recorded back to lineage ledger — full loop closed | v9.11.0 |
Quick Start
Desktop / Linux / macOS / CI
git clone https://github.com/InnovativeAI-adaad/ADAAD.git && cd ADAAD
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.server.txt
PYTHONPATH=. python -m pytest tests/test_boot_preflight.py -q # always first
ADAAD_CEL_ENABLED=true ADAAD_SANDBOX_ONLY=true ADAAD_ENV=dev python app/main.py
Android — Termux / Pydroid3
pip install -r requirements.phone.txt
CRYOVANT_DEV_MODE=1 ADAAD_ENV=dev python app/main.py --host 0.0.0.0 --port 8000
Aponi Governance Console
PYTHONPATH=. python -m ui.aponi --port 8080 # → http://localhost:8080
🟢 Live System Status
| System | Status |
|---|---|
| 🟢 GovernanceGate | ACTIVE · NON-BYPASSABLE · GOV-SOLE-0 |
| 🟢 Constitutional Evolution Loop | 14-STEP · ORDERED · CEL-ORDER-0 |
| 🟢 Evidence Ledger | HASH-CHAINED · APPEND-ONLY · CEL-EVIDENCE-0 · 12,441 entries |
| 🟢 FitnessEngine v2 | 7 SIGNALS · ADAPTIVE WEIGHTS · FIT-BOUND-0 |
| 🟢 SandboxTournament | OPERATIONAL · EPHEMERAL · SANDBOX-DIV-0 |
| 🟢 Deterministic Replay | ZERO DIVERGENCE · DET-ALL-0 |
| 🟢 WeightAdaptor | EMA DESCENT · TELEMETRY LIVE · FIT-DIV-0 |
| 🟢 Federation | HMAC-GATED · DUAL-GATE · federation_dual_gate |
| 🟢 CodeIntelModel | SCANNING · DETERMINISM-VERIFIED · INTEL-DET-0 |
| 🟢 Innovations Pipeline | SEED-CEL-AUDIT-0 · SEED-REVIEW-0 · SEED-PROP-0 |
| 🟢 Autonomous Self-Evolution | PHASE 65 PROVEN · LEDGER SEALED · MUTATION-TARGET |
Architecture
Three LLM-backed agents — Architect, Dream, Beast — compete for epoch selection via UCB1 bandit. Every winning proposal traverses the full 14-step Constitutional Evolution Loop before a single byte changes. The GovernanceGate is the sole approval surface. Invariant GOV-SOLE-0: it cannot be bypassed. Ever.
The 14-Step Constitutional Evolution Loop
All 14 steps execute in strict sequence. One failure → clean halt. Zero silent errors. CEL-ORDER-0: the order is structural, not configurable.
| Step | Name | Type | On Failure |
|---|---|---|---|
01 |
MODEL-DRIFT-CHECK |
Guard | Blocks epoch — determinism state stale |
02 |
LINEAGE-SNAPSHOT |
Capture | Records capability_graph_before hash |
03 |
FITNESS-BASELINE |
Measure | Pre-epoch 7-signal composite recorded |
04 |
PROPOSAL-GENERATE |
Generate | Architect · Dream · Beast in parallel · seed advisory if injected |
05 |
AST-SCAN |
Preflight | StaticSafetyScanner — 4 hard AST rules |
06 |
SANDBOX-EXECUTE |
Test | Ephemeral clone — SANDBOX_ONLY respected |
07 |
REPLAY-VERIFY |
Verify | Hash mismatch → auto-rollback (SANDBOX-DIV-0) |
08 |
FITNESS-SCORE |
Score | Determinism divergence → unconditional veto |
09 |
GOVERNANCE-GATE-V2 |
Gate | 5 diff-aware AST rules — Class A / B split |
10 |
GOVERNANCE-GATE |
Gate | 23 constitutional rules — ALL must pass |
11 |
LINEAGE-REGISTER |
Register | Survivors chained into lineage DAG |
12 |
PROMOTION-DECISION |
Promote | CapabilityGraph + PromotionEvent recorded |
13 |
EPOCH-EVIDENCE |
Seal | SHA-256 hash-chained ledger entry — immutable |
14 |
STATE-ADVANCE |
Advance | Epoch counter + epoch_complete.v1 emitted |
ADAAD_SANDBOX_ONLY=true → all 14 steps run · zero writes · full audit trail
Step 4 Seed Advisory (Phase 76)
Since Phase 76, an approved seed ProposalRequest may be injected into CEL Step 4 via inject_seed_proposal_into_context(). The seed is advisory only — Step 4 falls back to the default ProposalRequest if the key is absent. SEED-CEL-HUMAN-0: injected context cannot trigger mutation without GovernanceGate + HUMAN-0 sign-off.
# CEL Step 4 — Phase 76 seed advisory wire
context = inject_seed_proposal_into_context(context, seed_entry, ledger)
request = resolve_step4_request(context) # seed if present, default if absent
Three Competing Mutation Agents
|
Architect Governance-safe structural design. Low risk (0.1–0.4). Owns Tier-0 guard decisions and constitutional compliance. |
Dream High-novelty exploration. High expected gain (0.5–0.9). Sees 200 epochs ahead. Experimental and behavioral mutations. |
Beast Mutation velocity at the constitutional edge. Highest risk, highest pressure. Drives fitness regression targets hard. |
All three compete via UCB1 multi-armed bandit — switches to Thompson Sampling after ≥30 non-stationary epochs. Up to 12 candidates per epoch via BLX-alpha crossover. The GovernanceGate decides. The bandit learns.
UCB1 selection — values adapt each epoch
─────────────────────────────────────────────────────────────────
Architect ██████████ UCB 0.73 risk 0.2 EXPLOIT
Dream ███████████████ UCB 0.81 risk 0.6 EXPLORE ◄ selected
Beast ██████████████ UCB 0.79 risk 0.9 EXPLOIT
─────────────────────────────────────────────────────────────────
Winning proposal → 14-step CEL → GovernanceGate → ledger seal
The 23 Constitutional Rules
| 🔴 9 Blocking | 🟡 5 Warning | 🔵 9 Advisory |
|---|---|---|
| Unconditional halt | Tier-gated escalation | Audit trail only |
| No override. No exceptions. | Blocking in Tier 0 | Recorded every epoch |
🔴 9 Blocking Rules — Unconditional halt. No override. No exceptions.
| Rule | Enforcement |
|---|---|
single_file_scope |
Mutations confined to one target file |
ast_validity |
Resulting AST must parse cleanly |
no_banned_tokens |
No exec, eval, os.system, shell equivalents |
signature_required |
Every mutation must carry a valid cryptographic signature |
lineage_continuity |
Must chain to an existing, verified lineage record |
resource_bounds |
Memory and compute bounds enforced per tier |
federation_dual_gate |
Cross-repo: GovernanceGate approval in both repos |
federation_hmac_required |
HMAC validation on all federation channel ops |
soulbound_privacy_invariant |
No identity-linked data exported outside governed channels |
🟡 5 Warning Rules — Flagged in Tier 1/2. Blocking in Tier 0.
| Rule | Threshold |
|---|---|
max_complexity_delta |
Cyclomatic complexity delta ≤ +2 |
test_coverage_maintained |
Coverage delta ≥ 0 — must never decrease |
max_mutation_rate |
Epoch mutation density within anti-flood rate |
import_smoke_test |
All new imports resolve at import time |
entropy_budget_limit |
Proposal entropy within deterministic bounds |
🔵 9 Advisory Rules — Recorded in every audit trail and evidence bundle.
| Rule | Tracks |
|---|---|
deployment_authority_tier |
Tier-based deployment authority |
revenue_credit_floor |
Credit floor for contributor attribution |
reviewer_calibration |
Historical reviewer calibration score |
bandit_arm_integrity_invariant |
UCB1 / Thompson Sampling state consistency |
market_signal_integrity_invariant |
Market signal provenance and freshness |
gate_v2_complexity |
GovernanceGateV2 AST complexity diff |
gate_v2_import_surface |
GovernanceGateV2 import surface delta |
gate_v2_error_guard |
GovernanceGateV2 error guard preservation |
gate_v2_semantic_scope |
GovernanceGateV2 semantic scope adherence |
8 Permanent Architectural Invariants
Fitness Engine — 7 Signals
Every mutation is scored before GovernanceGate sees it. Determinism divergence is an unconditional veto — it cannot be overridden by any weight or governance exception.
| Signal | Weight Range | Veto |
|---|---|---|
| 🧪 Test delta — coverage + failure reduction | [0.05, 0.70] |
|
| 🌀 Complexity delta — cyclomatic change | [0.05, 0.70] |
|
| ⚡ Performance delta — runtime benchmark | [0.05, 0.70] |
|
| 📊 Governance compliance — 10-epoch rolling | [0.05, 0.70] |
|
| 🏗 Architectural fitness — coupling + centrality | [0.05, 0.70] |
|
| 🔒 Determinism — replay divergence | [0.05, 0.70] |
ANY divergence = halt |
| 🧹 Node economy — net AST additions | [0.05, 0.70] |
score = sum(signal_value[i] * adaptive_weight[i] for i in range(7))
if determinism_divergence > 0: raise VetoError("FIT-DIV-0") # unconditional
if score < baseline_composite: raise RejectError("below baseline")
if not gate_v2.approve(patch): raise RejectError("GovernanceGateV2")
if not gate.approve(patch): raise RejectError("GovernanceGate")
# → APPROVED
Mutation Tier System
| Tier | Paths | Mutations | Review |
|---|---|---|---|
| 🔴 0 | runtime/, security/, app/main.py |
❌ Never auto-applied | Human required |
| 🟡 1 | tests/, docs/, most agents |
✅ Auto-execute | Within 24h |
| 🟢 2 | app/agents/test_subject/ |
✅ Fully autonomous | Logs monitored |
TIER0-SELF-0 is permanent and structural. Tier-0 modules cannot be mutation targets — not via configuration, not via exception token, not via governance override.
Innovations & Seed Pipeline (Phases 69–75)
The Innovations Pipeline is ADAAD's governed pathway for human-sourced ideas to enter the constitutional evolution loop. Seeds are ingested, scored, promoted, reviewed by a human operator, converted to ProposalRequest objects, and injected as advisory signals into CEL Step 4.
Human Idea
│
▼ POST /innovations/seeds
SeedIngestion ─── SEED-INGEST-0 ──► LineageLedger
│
▼ automatic scoring
SeedPromotion ─── SEED-PROMO-0 ──► PromotionQueue
│
▼ POST /seeds/promoted/{id}/review (audit:write · HUMAN-0)
SeedReview ─── SEED-REVIEW-0 ──► SeedReviewDecisionEvent
│ approved only
▼ POST /seeds/promoted/{id}/propose
SeedBridge ─── SEED-PROP-0 ──► ProposalRequest (advisory)
│
▼ POST /seeds/promoted/{id}/inject
CEL Step 4 ─── SEED-CEL-AUDIT-0 ──► resolve_step4_request()
│ ↓ GovernanceGate
└─────────────────────────────────────► Evidence Ledger
All 6 seed invariants (SEED-CEL-0, SEED-CEL-HUMAN-0, SEED-CEL-DETERM-0, SEED-CEL-AUDIT-0, SEED-PROP-0, SEED-REVIEW-0) are enforced at each stage. A seed cannot trigger a mutation without full GovernanceGate + HUMAN-0 approval. No exceptions.
Federation — Cross-Repo Governed Evolution
Mutations propagate across repositories via FederationMutationBroker (Phase 5). Every federated mutation requires GovernanceGate approval in both repos, HMAC key validation, and zero cross-repo divergence in FederatedEvidenceMatrix.
Source Repo Federation Channel Destination Repo
─────────── ───────────────── ────────────────
ProposalEngine → HMAC envelope → ProposalTransport
GovernanceGate ─────────────────────────► GovernanceGate
EvidenceMatrix ◄─── divergence == 0 ────► EvidenceMatrix
EvolutionLedger EvolutionLedger
federation_dual_gate is a blocking constitutional rule. Bypassing either gate is structurally impossible.
The Aponi Dashboard
Aponi is the human oversight surface — epoch evidence review, lineage health monitoring, HUMAN-0 gate exercise, and (Phase 73+) the Seed Review panel.
┌──────────────────────────────────────────────────────────────────┐
│ APONI · Governance Console · ADAAD v9.11.0 · Phase 76 │
├──────────────────────────────────────────────────────────────────┤
│ Evolution Loop ● ACTIVE Epoch 1,247 │
│ GovernanceGate ● ENFORCING 23 rules / 0 bypass │
│ Evidence Ledger ● LIVE 12,441 entries │
│ Active Agents ● 3/3 Architect Dream Beast │
│ Innovations ● ACTIVE Seeds: 12 pending / 3 approved │
├──────────────────────────────────────────────────────────────────┤
│ Last Epoch APPROVED Class A Δ+2 tests │
│ Fitness 0.74 above baseline ✓ │
│ Replay Hash a3f8e9… divergences: 0 │
│ WeightAdaptor pred_accuracy 0.71 EMA trending ↑ │
│ Seed CEL advisory injected → Step 4 resolved ✓ │
└──────────────────────────────────────────────────────────────────┘
PYTHONPATH=. python -m ui.aponi --port 8080 # → http://localhost:8080
DEVADAAD — The Governed Build Agent
Use ADAAD or DEVADAAD as trigger tokens with any LLM following AGENTS.md:
| Trigger | Effect |
|---|---|
ADAAD |
Governed build — stages work for review, no merge |
DEVADAAD |
All ADAAD constraints + operator-authorized merge |
ADAAD status |
Orientation report — no build action |
ADAAD preflight |
Preflight checks only |
ADAAD audit |
Open findings from .adaad_agent_state.json |
ADAAD verify |
Full verify stack — no new code |
ADAAD retry |
Retry last blocked step post-remediation |
DEVADAADgrants merge authority. First word must beDEVADAADexactly. All 23 constitutional rules apply identically.
python -m app.main --adaad-status --trigger-mode ADAAD --status-format both
Trigger mode : ADAAD
Next PR : PR-76-PLAN
Dependency readiness : READY
Tier 0: pass | Tier 1: pass | Tier 2: pass | Tier 3: pass
Pending evidence rows: none
Install on Any Platform
| Platform | Install | Notes |
|---|---|---|
| F-Droid | f-droid.org | Android · open source · no tracking |
| Obtainium | obtainium.imranr.dev | Android · auto-update from GitHub |
| GitHub Releases | releases | All platforms · Linux · macOS · Windows WSL |
| PWA | install page | Browser · any device · no install |
Codebase at a Glance
| Metric | Value |
|---|---|
| Version | 9.10.0 — Phase 76: Seed CEL Outcome Recorder |
| Runtime Python | 329+ files |
| Test files | 451 · ~40,000+ lines |
| Passing tests | 4,624 |
| Constitutional rules | 23 — Constitution v0.9.0 |
| Architectural invariants | 8 — code-level, non-overridable |
| Seed invariants | 6 — SEED-CEL-0 through SEED-REVIEW-0 |
| Development phases | 76 complete |
| Evidence ledger entries | 12,441 — SHA-256 hash-chained |
| Android support | Full — Termux + Pydroid3 |
| Federation | Multi-repo HMAC-gated (Phase 5) |
| First self-evolution | ✅ Phase 65 — March 13, 2026 |
| Proof | SHA-256 hash-chained, deterministically replayable |
Governance Authority Chain
╔══════════════════════════════════════════════════════════════╗
║ docs/CONSTITUTION.md · v0.9.0 · 23 rules ║
║ Root of all governance authority ║
╚═════════════════════════╤════════════════════════════════════╝
│
▼
╔══════════════════════════════════════════════════════════════╗
║ docs/ARCHITECTURE_CONTRACT.md ║
║ 8 invariants · structural constraints · module bounds ║
╚═════════════════════════╤════════════════════════════════════╝
│
▼
╔══════════════════════════════════════════════════════════════╗
║ docs/governance/ARCHITECT_SPEC_v3.1.0.md ║
║ Implementation spec · organ contracts · enforcement ║
╚═════════════════════════╤════════════════════════════════════╝
│
▼
┌──────────────────────┐
│ GovernanceGate │ ◄── GOV-SOLE-0
│ 23 rules active │ Non-bypassable
│ Every mutation │ Not configurable
└──────────┬───────────┘
│ APPROVED
┌─────────────────┼─────────────────┐
▼ ▼ ▼
EvolutionLedger CapabilityGraph LineageLedger
SHA-256 chain version graph stability DAG
Project Structure
ADAAD/
├── app/ # Orchestrator · agents · mutation cycle · FastAPI
│ └── agents/ # Architect / Dream / Beast
├── runtime/ # Core engine — the constitutional machine
│ ├── evolution/ # CEL · lineage · fitness · replay verifier
│ ├── governance/ # GovernanceGate · federation · rate limiting
│ ├── autonomy/ # Bandit · AdaptiveWeights · NonStationarityDetector
│ ├── mutation/ # AST substrate · SandboxTournament · CodeIntelModel
│ ├── sandbox/ # Ephemeral container execution
│ ├── seed_cel_injector.py # Phase 76 — seed → CEL advisory wire
│ ├── seed_proposal_bridge.py # Phase 74 — seed → ProposalRequest bridge
│ └── seed_review.py # Phase 73 — human-governed seed review
├── security/ # Cryovant — auth · key management
├── tests/ # 451 files · 4,624 passing
├── governance/ # Constitutional rules · federation keys · attestations
├── artifacts/ # Per-phase evidence (immutable after close)
├── ui/ # Aponi governance console
│ └── aponi/ # innovations_panel.js · Phase 73/74/75 seed UI
├── android/ # F-Droid · Obtainium · GitHub Releases · PWA
├── QUICKSTART.md
├── AGENTS.md # ADAAD / DEVADAAD build agent protocol
├── ROADMAP.md
└── docs/CONSTITUTION.md # The 23 rules — root of all authority
FAQ
Is ADAAD safe to run on a production codebase?
Start with
ADAAD_SANDBOX_ONLY=true ADAAD_ENV=dev. The full 14-step CEL executes with zero writes. When ready, use Tier 2 (sandbox paths) for first live runs. Tier 0 (production runtime) requires human sign-off. TIER0-SELF-0: core modules cannot be self-mutated under any condition.
What is Class A vs Class B?
GovernanceGateV2 classifies by AST complexity delta. Class A (delta ≤ +2): auto-approved if all other rules pass. Class B (delta > +2): requires an exception token and a
HUMAN-0 co-sign. Complex mutations escalate — they don't get blocked, they get governed with extra scrutiny.
ADAAD vs DEVADAAD?
Both are trigger tokens for the governed build agent in
AGENTS.md. ADAAD builds and stages for review — no merge. DEVADAAD adds operator-authorized merge. All 23 constitutional rules apply identically. The first word of the prompt must be DEVADAAD exactly.
What changed in v9.11.0 (Phase 76)?
Phase 76 wired approved seed
ProposalRequest objects into CEL Step 4 as advisory signals. New: runtime/seed_cel_injector.py · inject_seed_proposal_into_context() · resolve_step4_request() · POST /innovations/seeds/promoted/{id}/inject. Four new invariants: SEED-CEL-0, SEED-CEL-HUMAN-0, SEED-CEL-DETERM-0, SEED-CEL-AUDIT-0. 14 new tests. Zero regressions. Seed remains advisory — no mutation without GovernanceGate + HUMAN-0.
What is the Innovations Pipeline?
Phases 69–75 built a governed pathway for human-sourced ideas (seeds) to enter the constitutional evolution loop. Seeds flow: ingest → score → promote → human review (
audit:write scope) → proposal bridge → CEL Step 4 advisory injection. Every step writes to the lineage ledger before any state mutation. All seed invariants are enforced. A seed cannot become a mutation without full GovernanceGate + HUMAN-0 approval.
What changed in v9.1.0 through v9.9.0?
| Version | Phase | Key change |
|---|---|---|
| v9.1.0 | 66 | WeightAdaptor telemetry · LineageLedgerV2 invariants · LLM failover contract · Ed25519 key ceremony |
| v9.2.0 | 67 | Operator-gated deployment scope · governance contracts |
| v9.3.0 | 68 | LLM failover governance · provider health monitoring |
| v9.4.0 | 69 | Innovations ingestion router · seed scoring framework |
| v9.5.0 | 70 | Innovation scoring · POST /innovations/seeds pipeline |
| v9.6.0 | 71 | Seed promotion queue · SEED-PROMO-0 invariant |
| v9.7.0 | 72 | Graduated seed tracking · promotion lifecycle ledger |
| v9.8.0 | 73 | Human-governed seed review · audit:write scope · Aponi review panel |
| v9.9.0 | 74 | Seed-to-Proposal Bridge · deterministic cycle_id · SEED-PROP-0 |
| v9.11.0 | 75 | Seed CEL injection · resolve_step4_request() · SEED-CEL-AUDIT-0 |
What ADAAD Is Not
Not a general-purpose LLM coding assistant. Not an unattended auto-merge system. Not a CI/CD replacement. Not a model training framework. Not a tool where the human is the safety layer.
It is a governed code evolution engine. Safety is structural. The GovernanceGate is not a guardrail — it is the only path through. GOV-SOLE-0. Architectural. Not configurable.
License
Source code: Apache License 2.0 — free forever. Brand assets: Proprietary — see BRAND_LICENSE.md. Trademarks: "InnovativeAI", "ADAAD", "Aponi", "Cryovant" — InnovativeAI LLC.
Built by Dustin L. Reid, InnovativeAI LLC — Blackwell, Oklahoma.
⚡ Quickstart · 📜 Constitution · 🗺 Roadmap · 🤖 Agents · 🐛 Issues · 🤝 Contributing · 📱 Android · 📋 Changelog
ADAAD v9.11.0 · Phase 76 · Apache 2.0 · InnovativeAI LLC · Blackwell, Oklahoma · github.com/InnovativeAI-adaad/ADAAD
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file adaad-9.11.0.tar.gz.
File metadata
- Download URL: adaad-9.11.0.tar.gz
- Upload date:
- Size: 1.1 MB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.10.12
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
4ce00d581191090f979fa1df1d1d7ce8d329a2f1ee17e703ad839a94ed527209
|
|
| MD5 |
d2447d0b446a3c5bf0fe614581d6a21d
|
|
| BLAKE2b-256 |
9e21985c6970680c58632925d5da812cf72d7a006ec6beb083afa618e20e815b
|
File details
Details for the file adaad-9.11.0-py3-none-any.whl.
File metadata
- Download URL: adaad-9.11.0-py3-none-any.whl
- Upload date:
- Size: 1.0 MB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.10.12
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
b033c62d63a443db54f657aa629d346913966f0d624511e49711e3fe6c356987
|
|
| MD5 |
f47f8b81a9d2b99c5be39b1b10ed0a02
|
|
| BLAKE2b-256 |
8340773618bb0979f9e3d8e73694d839948546ec219da78a4cea83d9522ff90a
|