okto-pulse-core 0.2.3

Okto Pulse Core — spec-driven project management engine with MCP support for AI agents

okto-pulse-core

Core engine for Okto Pulse — shared models, services, API routes, and MCP server.

Ship with AI. Stay in control.

You probably want to install okto-pulse instead. This package is the internal engine. The okto-pulse package provides the CLI, frontend, and everything you need to get started.

What's inside

• 52 SQLAlchemy models — Boards, Cards, Specs, Ideations, Refinements, Sprints, Agents, Knowledge, Mockups, Validations, KG queues, rebuild/cognitive-candidate records, discovery entities and audit/outbox records. (Skills entity dropped in 0.2.0.)
• 28 service classes — Full business logic with governance rules, board agent governance, resource propagation + lineage, bug-regression workflow, archive/restore, traceability and board-level resource readiness. (Skills service dropped in 0.2.0.)
• 33 API route modules — FastAPI REST endpoints
• 17 governance gates — Resource readiness, resource-to-task coverage, spec coverage, validation, evaluation, task completion, cognitive closeout, architecture-findings, evidence, bug traceability and sprint health controls.
• 215 MCP tools — Complete Model Context Protocol server for AI agent integration, including:
  • Pipeline CRUD (Ideation, Refinement, Spec, Sprint, Card)
  • Q&A and choice questions across every entity
  • Mockups (HTML+Tailwind, sanitised) and Knowledge Bases at spec/refinement/card scope
  • Decisions with supersedence and coverage gates
  • Per-card Knowledge attachment lifecycle (add_card_knowledge and friends)
  • 24 Knowledge Graph tools (consolidation, query primary/power, health, dead-letter, schema-migrate, decay tick controllability, rebuild preflight/confirm/run)
  • Community runtime exposure: 215 core MCP tools, 0 community-only MCP tools
• App factory — create_app() with dependency injection for auth and storage providers
• Embedded Knowledge Graph — per-board LadybugDB instance + global discovery meta-graph, deterministic + cognitive workers, 11 node types and 10 relationship types

Governance Gate Surface

Okto Pulse currently documents and enforces 17 named governance gates:

Gate family	Gates
Resource readiness	Resource readiness; resource-to-task coverage
Spec coverage	Scenario/test coverage; functional requirement/business rule coverage; technical requirement/task coverage; API contract/task coverage; active decision/task coverage
Validation and evaluation	Spec validation; spec qualitative evaluation; task validation
Execution quality	Task start/spec readiness; task conclusion; cognitive closeout; architecture-findings done; test evidence; bug test-first/traceability
Sprint health	Sprint closure/evaluation

The two execution-quality additions in 0.2.3 — cognitive closeout (a done transition is blocked while active cognitive-consolidation items remain) and the architecture-findings done gate (active architecture warnings block spec/card done) — moved from defined to enforced in this release.

Architecture

┌─────────────────────────────────────────────────────────┐
│  Single Python process  (single Kùzu lock holder)        │
│                                                          │
│   ┌─────────────────┐         ┌──────────────────┐       │
│   │  uvicorn :api   │         │  uvicorn :mcp    │       │
│   │   FastAPI app   │         │  MCP ASGI app    │       │
│   │   /api/v1/*     │         │  /mcp            │       │
│   │   /static, SPA  │         │  streamable-http │       │
│   └────────┬────────┘         └────────┬─────────┘       │
│            │                           │                  │
│            └─────────┬─────────────────┘                  │
│                      │                                    │
│           shared module-level state:                      │
│           - Mongo session factory (init via lifespan)     │
│           - _global_db (Kùzu cache)                       │
│           - _active_api_key (ContextVar)                  │
└─────────────────────────────────────────────────────────┘

build_mcp_asgi_app() and mount_mcp(app) are the two helpers exposed from okto_pulse.core.mcp — pick build_mcp_asgi_app() to drive a separate uvicorn Server (the community edition does this for the --mcp-port listener) or mount_mcp(app) to mount the MCP sub-app under an arbitrary path on an existing FastAPI app.

Docker

This repo has no Dockerfile. The deployable artifact is a single image built from the sibling okto-pulse repo:

• okto-pulse/Dockerfile target local-runtime builds wheels from this repo and okto-pulse/ as siblings (used by okto-pulse/docker-compose.yml and the release pipeline's smoke build).
• okto-pulse/Dockerfile target pypi-runtime installs okto-pulse==<version> from PyPI, which transitively pulls this package off PyPI per the floor in okto-pulse/pyproject.toml (used by okto-pulse/docker-compose.prod.yml).

To run the published image:

docker run -d --name okto-pulse \
  -e HOST=0.0.0.0 -e MCP_HOST=0.0.0.0 \
  -p 8100:8100 -p 8101:8101 \
  -v okto-pulse-data:/data \
  ghcr.io/oktolabsai/okto-pulse:latest

See okto-pulse/README.md for the full Docker quickstart, and okto-pulse/CLAUDE.md for the multi-stage build architecture.

Release Notes

0.2.3 — current

The largest release since 0.2.0. Scope is taken from the 53 finalized specs on the Okto Pulse 0.2.3 board (the platform dogfooded its own SDLC), landing 64 new core modules across eight subsystems. 335 files changed, +103,183 / −4,532 over 0.2.2; every subsystem ships with its pytest suite. The package grew to 52 models / 28 services / 33 API modules / 215 MCP tools / 17 named gates.

• KG corruption prevention & durability (headline — KG-01, KGDL.01) — new write-path primitives (safe_write_lifecycle, write_barrier, single_writer_lock, backpressure, quarantine, contingency). A non-destructive durability lifecycle (STEP_CHECKPOINT no longer closes the cached Database; a per-board _BoardCloseGuard drains live readers before any close) eliminates the use-after-close of the shared Database — the probable second corruption vector for graph.lbug. Spec 3d89c192.
• KG recovery, reset & deterministic rebuild (KG-02 + R2a) — rebuild a board's graph from canonical SQL sources, deterministically and audited: rebuild_preflight/_confirm/_run via REST + agent-actionable MCP twins (confirmation-token gated, quarantine-aware), plus auto-recovery of interrupted checkpoints.
• KG zero-orphan integrity (KG-ZO-01/02) — a node-connectivity pre-commit guard that refuses to commit orphans, plus orphan backfill, health reporting and rebuild visibility.
• KG cognitive consolidation & source governance (KG-03/03A) — cognitive item control + candidate-decision promotion (candidate_decision_store, cognitive_badge_resolver), per-concept source_ref, and dedup granularity with SUPERSEDE wiring + counted/audited merge.
• KG health honesty & degraded-mode resilience (F3/F4/F16/F17, R2c) — signal clarity (scheduler/decay debt ≠ corruption; footprint = file-size proxy), a resilient/observable decay tick, a uniform graph_unavailable envelope, a health-aware closeout gate + tick admission, real memory-pressure instrumentation and opt-in DLQ auto-drain.
• Governance, lineage & gates (BG-01, RG-01, AFG) — critical_context_guard (critical mutations resolve + fingerprint full entity context first), the resource_lineage provenance resolver with N/A inheritance, and the Architecture Finding Done Gate wired into spec → done. Two gates moved from defined to enforced (15 → 17): Cognitive Closeout and Architecture Findings.
• MCP token-optimization & projection (R1–R5) — payload_budget, payload_compaction, projection_envelope, context/copy projection, kg_query_safety, tool_family_registry; schema honesty (anyOf array|string), positional → canonical id/ref migration (linked_requirements → FR, linked_criteria → AC), and the pre-flight checklist as a real okto-pulse://workflows/preflight resource. Surface grew to 215 tools.
• Bug-regression workflow — scenario reuse + test-gate remediation, operator-facing bug guidance/error remediation, and a post-closure hotfix lane.
• Structured spec entities + API-contract hardening — structured editing + test_scenario CRUD (closes the NC-9 bypass), contract_type discriminator + HTTP-method enum, granular per-requirement N/A for IR/OR/contract, and structured choice fields.
• Analytics & telemetry — IR/OR coverage calculator with a cancelled-card filter, Decision-coverage surfacing, and beacon-off metrics modes.

See CHANGELOG.md for the per-subsystem diff-level rationale.

0.2.2

Patch release rolling up the post-0.2.1 fixes. Same surface as 0.2.1 plus:

• SDLC E2E gate polish (4 issues from the 2026-05-17 ceremonial run) —
  • submit_spec_validation now runs the AC → test-scenario coverage check as its first pre-requisite, so a spec with uncovered ACs fails BEFORE the validation locks the content (previously the move → done gate raised the same error but only after the spec was already locked).
  • The "FR has no linked business rule" error now uses an [i] index marker rather than the duplicated FR{i}: prefix that collided with the author's own FRN: label and produced strings like "FR1: FR2: ...".
  • okto_pulse_link_task target_type='decision' now spreads the saturation envelope into its success JSON, in parity with the other six target types. A parametrised dispatcher test pins the contract so a future eighth helper can't regress it silently.
  • okto_pulse_evaluate_ideation docstring now states the status='evaluating' pre-requisite and the full draft → review → approved → evaluating → done flow up front. The tool deliberately does not auto-promote — each transition is an explicit gate decision.
• Agent instructions split + reference catalogue — agent_instructions.md trimmed by extracting the static reference material into three new MCP resources (okto-pulse://reference/list_tools, tools_catalog, transitions). Workflow docs (refinements, specs, stories) refreshed in lockstep.
• MCP server module slim-down — core/mcp/server.py lost ~970 lines of helpers that now live in supporting modules. Public tool surface is unchanged; this is purely an organisational refactor on top of 0.2.1.

Anti-regression tests added for each of the four E2E fixes (tests/test_spec_validation_gate.py::TestAcScenarioPrecheck, TestFrCoverageMessageFormat, tests/test_link_task_dispatcher.py::test_link_helper_returns_saturation_envelope). The submit_spec_validation baseline hash in tests/.cache/validation_gates_baseline.txt was bumped to reflect the intentional addition of the new AC → scenario pre-check.

See CHANGELOG.md for the diff-level rationale and the per-fix bug card references.

0.2.1

Branch changelog (feature/0.2.1)

This branch turns 0.2.1 into the IR/OR, telemetry, resource-propagation and MCP-surface optimization release.

• Added local-first telemetry and metrics infrastructure: product event schema, settings model, local event store, sender, privacy-aware service layer, metrics REST API and tests for local-only, disabled and anonymous-beacon modes.
• Added first-class Integration Requirements (IR) and Observability Requirements (OR) across database models, Pydantic schemas, REST responses, MCP handlers, permission registry and presets.
• Extended spec context, sprint context and coverage summaries so agents and UI callers can see IR/OR items alongside technical requirements, business rules, API contracts, decisions and test scenarios.
• Added service-layer spec resource propagation. Knowledge Base entries, architecture designs and mockups can be copied from specs to cards automatically when board settings enable auto-derive resources.
• Hardened propagation triggers on card creation/linking, spec resource edits and architecture updates so downstream task cards stay self-contained without relying on "see the spec" references.
• Added granular IR/OR permissions (read, create, link_task) and enforced them consistently across API, MCP and permission presets.
• Added four consolidated MCP list handlers: okto_pulse_list_by_board, okto_pulse_list_qa, okto_pulse_list_knowledge and okto_pulse_list_snapshots.
• Added server-side MCP filter validation and JSON-string filter decoding so tool transports can pass either dict objects or JSON-encoded filter strings.
• Split the large root MCP agent instructions into a compact pre-flight plus 12 lazy MCP resources under okto-pulse://workflows/... and okto-pulse://reference/....
• Added a runtime MCP schema-generation pilot for card CRUD tools, backed by Pydantic v2 model schemas and a snapshot fixture.
• Added minimal-envelope response modes and token-optimization refinements for agent workflows, including tighter list/context payloads and refreshed workflow documentation.
• Added cursor-based keyset pagination to okto_pulse_get_activity_log, including opaque next_cursor, invalid-cursor structured errors and SQLite timestamp normalization for microsecond-safe pagination.
• Improved activity-log summaries and card-move logging by covering more action shapes and de-duplicating noisy card_moved entries.
• Hardened architecture services with semantic normalization and additional validation coverage used by the community Architecture UI and Excalidraw import flow.
• Added focused regression coverage for spec resource propagation, telemetry, IR/OR requirements, consolidated MCP handlers, MCP resources, schema generation, activity-log pagination and story/refinement regressions.

Post-release polish already on the branch

• Aligned authoritative handler signatures for the four consolidated MCP list handlers.
• Unified list handler defaults to limit=100 for consistency across old and consolidated paths.
• Clarified that the implementation keeps board_id for ACL/auth and uses entity_type/entity_id naming for knowledge listings.

SDLC E2E gate polish (4 issues from end-to-end run 2026-05-17)

A full ceremonial E2E run (Story → Ideation → Refinement → Spec → Sprint → Cards → Sprint closeout) on the E2E board surfaced four small but recurring issues across the spec validation gates, error messages and tool response shapes. All four were fixed in the same Unreleased cycle and validated in-vivo against the live MCP server. See CHANGELOG.md for the full diff and rationale; the short summary:

• submit_spec_validation now runs the AC → test-scenario coverage check as the first pre-requisite, so a spec with uncovered ACs fails BEFORE the validation locks it. The error message also reminds the caller that the spec is locked after a successful validation.
• The "FR has no linked business rule" error message now uses an [i] index marker instead of FR{i}:, removing the confusing FR1: FR2: ... duplication that occurred whenever the FR text already started with its own label.
• okto_pulse_link_task with target_type='decision' now returns the same saturation envelope as the other six target types. Previously only the decision branch returned the bare {success, decision_id, card_id, linked_tasks} shape, breaking agents that drive "continue linking vs submit validation" off the saturation signal.
• The okto_pulse_evaluate_ideation MCP docstring now states the status='evaluating' pre-requisite and the full draft → review → approved → evaluating → done flow up front, so agents stop discovering the requirement by trial and error.

Anti-regression tests were added for each fix (test_spec_validation_gate.py::TestAcScenarioPrecheck, TestFrCoverageMessageFormat, and test_link_task_dispatcher.py::test_link_helper_returns_saturation_envelope). The submit_spec_validation baseline hash in tests/.cache/validation_gates_baseline.txt was bumped to reflect the intentional addition of the new coverage call.

0.2.0

Branch changelog (feature/0.2.0)

This branch turns 0.2.0 into the governed SDLC + Knowledge Graph release.

• Added Stories and Topics as pre-ideation intake primitives, including REST/MCP services, permissions, lifecycle rules, story-to-ideation traceability and the rule that a Story can reference only one Ideation while an Ideation can reference many Stories.
• Added Resource Gate readiness across Architecture, Mockups and Knowledge Base, with reversible N/A justification, entity-level readiness summaries and MCP guardrails that keep deterministic resource checks out of ad-hoc agent judgement.
• Hardened agent instructions for ambiguity handling: agents are directed to ask more clarification questions, prefer multiple-choice questions with recommendations when possible, and preserve an additional comment path for user nuance.
• Added Ideation Knowledge Base support and propagation, plus lineage/reporting improvements so specs, sprints, tasks, tests and bugs remain traceable even when a flow intentionally starts at Spec without a root Ideation.
• Expanded deterministic KG ingestion for specs, cards, bugs, tests, outcomes, requirements, criteria, constraints, API contracts and decisions, including resolved Bug originates_from and covered_by edges and schema migration coverage for those relationship tables.
• Strengthened KG schema lifecycle and graph runtime resilience: per-board schema bootstrap/migration, edge metadata migration, entity dedup support, Kuzu memory/runtime settings, vector-extension loading on hot-path graph connections and richer health/dead-letter diagnostics.
• Improved KG query/display contracts: /kg/boards/{board_id}/graph now accepts a node type filter, /nodes total hints remain filter-aware, graph stats expose node/edge histograms and tests cover pagination, type filtering and schema edge counts.
• Fixed guideline creation/parsing paths that could reject inline guideline additions with 422 responses.
• Preserved test scenario evidence in REST response schemas, including latest_evidence fallback data, so UI audit surfaces can expose recorded execution proof for Test cards.
• Added and expanded focused tests for Stories, Topic permissions, Resource Gate, Ideation KB, guidelines, deterministic KG workers, graph pagination, schema migration, traceability reports, presets and MCP registration contracts.

Fix C: single-process, dual-port serve (Kùzu lock contention)

okto-pulse serve now runs API/UI and MCP from a single Python process but on two different ports (--api-port defaults to 8100, --mcp-port defaults to 8101). Two uvicorn.Server instances run concurrently inside one asyncio.gather — the embedded Kùzu DB is owned by exactly one OS process (no inter-process lock contention), and the two listeners share the module-level state (the registered session factory, the _global_db Kùzu cache, the _active_api_key ContextVar).

What you get:

• No Kùzu file-lock thrash — the embedded DB does not support multiple writers, so a single Python process is the only safe topology. The kg.db_open.lock_retry path=... attempt=N/5 warnings disappear.
• Independent ports — keep :8100 for the SPA fetches and :8101 for the MCP HTTP transport, unchanged from earlier releases.
• One lifespan — init_db, KG worker startup, scheduler boot, and register_session_factory all run once on the API listener; the MCP sub-app picks up the registered factory automatically.

Public surface:

• okto_pulse.core.mcp.build_mcp_asgi_app() — returns the MCP ASGI app wrapped in the ApiKeySessionMiddleware (handles ?api_key= / X-API-Key / Authorization: Bearer and binds the key to the request ContextVar).
• okto_pulse.core.mcp.mount_mcp(app, mount_path="/mcp") — mounts the same ASGI app onto an existing FastAPI app at the given path (community does this when an embedded mount is needed).
• okto_pulse.core.mcp.register_session_factory(factory) — call from the API lifespan so the MCP sub-app finds the DB. Idempotent.

Spec Skills entity removed in its entirety

The experimental "skills" feature on the spec entity is gone. Adoption was zero in real boards and knowledge entries already cover the reusable-context use case more naturally — the dedicated tab, MCP tools, REST endpoints and ORM table were paying recurring maintenance cost without return.

What goes away:

• 5 MCP tools removed — okto_pulse_create_spec_skill, okto_pulse_delete_spec_skill, okto_pulse_spec_skill_retrieve, okto_pulse_spec_skill_inspect, okto_pulse_spec_skill_load.
• 4 REST endpoints removed — GET / POST / PATCH / DELETE /api/v1/specs/{spec_id}/skills (and the {skill_id} variants).
• 5 permission flags removed — spec.skills.{read,load,create,delete,recall} from the registry and from every preset.
• Database table dropped — spec_skills. Migration is idempotent (DROP TABLE IF EXISTS); no downgrade — the data is gone.
• Pydantic schemas removed — SkillSectionSchema, SpecSkillCreate, SpecSkillUpdate, SpecSkillResponse, SpecSkillSummary. The skills field is gone from SpecResponse.
• agent_instructions.md scrubbed — Quick Navigation, the dedicated Spec Skills section, the spec-authoring workflow step and the destructive-operations row no longer reference skills.

Reader-side defensive handling: BaseSchema now sets extra="ignore" so historical payloads still carrying a skills field validate silently — no warning, no log, no error. There is nothing to migrate; the field is dropped on read.

Use knowledge entries (spec_knowledge, card_knowledge) and decisions for the same use case.

Agent instructions overhaul

agent_instructions.md was reviewed end-to-end. Three behavioural sections were added in response to repeated drift patterns observed across production sessions:

• § 2.1a Ambiguity-killer protocol — at ideation, the agent must scan the user's request against a table of ambiguity symptoms (vague verbs, undefined nouns, multiple plausible interpretations, implicit success criteria, implicit scope) and post Q&A items for every gap before advancing the ideation. "Just make a reasonable choice" is permission, not silence — it must be recorded explicitly.
• § 2.2a Investigação profunda obrigatória (refinement) — refinement is research, not paraphrasing. The agent must exhaust all applicable sources (project files, source code, KE, Knowledge Graph, mockups, web docs, online discussions, runtime evidence, stakeholder context) and the refinement body must cite each finding with path:line, KE titles, KG node ids or URLs.
• § 2.8 Card-level artifact attachment (MANDATORY) — every card must be self-contained. KE/mockup dependencies must be attached directly to the card via copy_knowledge_to_card / copy_mockups_to_card / add_card_knowledge / add_screen_mockup(entity_type="card"). Vague references to "see the spec" are a protocol violation.

Cleanup:

• Quick Navigation header Multi-value Parameters — Two Accepted Formats corrected to Three Input Shapes (the section was extended to native list[str] in 0.1.4).
• Obsolete delete_task_validation reference removed from the Available Tools → Evaluations & Validations table (the tool never shipped).
• okto_pulse_create_sprint parameter list aligned with the schema (objective? and expected_outcome? were missing).
• Duplicate "Startup Protocol" subsection deleted — Pre-Flight Checklist is the single source of truth.

Other improvements

• MCP ApiKeySessionMiddleware rewritten on top of ContextVar — required because the FastAPI process serves multiple concurrent requests and the previous module-level global would leak identities across requests. Token-based set/reset pattern protects against exception leaks.
• run_mcp_server retained for legacy callers, now defers to build_mcp_asgi_app for ASGI-mode embeds.

To upgrade an existing install: pip install -U okto-pulse okto-pulse-core and then okto-pulse init --agents to regenerate .mcp.json (the URL still points at port 8101 by default; override with --mcp-port if you remapped). No downstream contract changes for MCP clients — the wire protocol and tool catalog (sans 5 skills tools) are unchanged.

0.1.3 — previous stable (PyPI)

First hardening pass on the card lifecycle, the analytics contract, and the MCP instruction set.

• CardService.delete_card cascades through every spec-side JSON list (test_scenarios[].linked_task_ids, business_rules[], api_contracts[], technical_requirements[], decisions[]) and through bug cards' linked_test_task_ids. The transactional cascade unblocks the delete→recreate flow that previously tripped _validate_spec_linked_refs.
• Analytics card-type classifier uses enum identity instead of str(card.card_type).endswith(...). total_cards_impl/test/bug, task_validation_gate.total_submitted, velocity[].test/bug, and bug_rate_per_spec now report real counts.
• parse_multi_value helper consolidated the scattered .split("|") pattern; pipe-separated and JSON-array inputs are autodetected.
• MCP agent instructions rewritten (1830 → 2050 lines) with new sections for Multi-value Parameters, Destructive Operations, Versioning & Concurrent Edits, Security, Analytics-Driven Closure.

0.1.1 — initial PyPI release

26+1 SQLAlchemy models, 17+1 service classes, 11 API route modules, 119 MCP tools, embedded Kùzu Knowledge Graph with deterministic workers. (Spec Skills shipped here and was removed in 0.2.0.)

(Version 0.1.2 was published to TestPyPI only as a release candidate for 0.1.3.)

License

Elastic License 2.0 — free for personal and commercial use. Cannot be offered as a hosted/managed service.