Cortex

August 28, 2026 · View on GitHub

Cortex is a DeepSeek-Harness plugin that turns the host's primary (strong) model into a high-leverage supervisor. The supervisor plans, decomposes, verifies and accepts; routed executor models (small/cheap models such as Qwen3.8-9B) and tools do the heavy lifting. With a recursive task tree, dynamic routing, four-stage quality gates, budget control, failure recovery, policy reuse and per-dimension model evaluation, Cortex keeps quality stable while driving the unit cost of successful tasks down.

The supervisor is the host primary model that drives the cortex_* tools (e.g. deepseek-v4-pro in the bundled registry), and executors are the models the dynamic router picks per subtask (L1/L2 tiers, e.g. Qwen3.8-9B).

Suggested repository topics (GitHub): dsh-plugin · dsh · cordis · llm-orchestration · multi-model · model-routing · quality-gate

Highlights

Capability	Entry	Notes
Task intelligence	`cortex_start`	Profile → fingerprint → strategy match (≥0.90 reuse / 0.70~0.90 light-check / <0.70 replan) → budget pool → granularity check
Recursive task tree	`cortex_decompose`	Safety limits: maxDepth 6 / maxTotalNodes 50 / maxChildrenPerNode 10 / maxReplan 3; DAG `depends_on`; aggregate nodes merge children by node schema + Gate0, active-model only
Recursive execution	`cortex_execute recursive`	Ready leaves auto-scheduled, aggregate nodes merged; evaluate:gate2 / multi_vote pass through — same standard as manual batch
Caching	engine built-in	L1 exact + L2 semantic (near-duplicate text similarity, conservative 0.88 threshold, skip-able) + L3 cross-task component reuse
Drift & value	engine built-in	Model drift monitoring vs static baseline (auto routing penalty, surfaced in registry/matrix) + value score & Pareto frontier per evaluation run
Dynamic routing	`cortex_execute`	Candidate filter → utility scoring (quality/success/risk − cost/latency/failure) → MQC upgrade gate; costFirst / balanced / qualityFirst; multi-model vote
Quality gates	Gate0 deterministic · Gate1 self-check · Gate2 evaluator · Gate3 `cortex_review` · Gate3.5 red-team	Hard fails: format_validity / safety / groundedness < 0.5; review decisions carry node quality/Gate1 context snapshots; red-team adversarially challenges high-impact acceptances (S/M, risk≥0.5 or quality≥0.9) — material defect → downgrade to retry
Recovery engine	`cortex_recover`	Failure taxonomy → retry / switch / prompt-fix / decompose / upgrade / human + failure-pattern learning
Soft-defect correction	`cortex_execute` / `cortex_recover`	Soft cap `max_delegations` (default 3): a capable-but-insufficient result is corrected in place (same model → better model → escalate to supervisor); optional continuable sub-session (`config.continuableCorrection`) keeps context across rounds; hard failures still use type-based recovery
Price override	`cortex_models` / UI / CLI	Set/clear a model's effective price (`price`/`clear_price`, or `/cortex models set-price
Budget control	engine built-in	4 pools (10/70/10/10), STOP-1..4 (STOP-2 from observed MQC — real ΔQ/ΔC, not hard-coded), 90% high-cost lock, degraded mode
Model evaluation	`cortex_evaluate`	7-dim scoring per case; per-model × task-type × capability × 7-dim profile (`dims_profile`, EMA, persisted across restarts; also fed by production Gate2/`cortex_review` dims); dynamic capabilities (profile-thresholded caps auto-register into the catalog and union with static capabilities at routing); Rule-First deterministic correctness for structured `expected`; UI one-click run; value score (Q×SR/Cost) + Pareto frontier; drift monitoring (vs static baseline → automatic routing penalty)
Model registry	`cortex_models` (`config/models.yaml`)	Static YAML + DSH `ctx.llm` dynamic discovery + enable/disable management; routing scores (ranking preview)
Policy center	`cortex_policy` / `cortex_strategy_match`	Fingerprint similarity, versioned policies, skip-level tree suggestions
KPI	`cortex_report`	Supervisor calls/task, A-cost breakdown by call kind (start/decompose/review/…) (aByKind), supervisor leverage, saving vs direct primary-model run (negative → orchestration warning), quality, policy reuse rate
Web console	`/cortex ui` / 8788	Trends / timeline / model registry / capability matrix (4 metrics + 7-dim rows) / evaluation center (case library + results), i18n zh+en, light/dark theme, CNY/USD display, SSE live updates
Watchdog	`/cortex watchdog` / `/watchdog`	Cross-workspace timed task injection + goal-reached auto-iteration: schedule (`--at/--after/--every`, every ≥300s) injected via the official inbox channel (`agent.followup` — never writes session events), goal-loop (set goals; after an injected/executed task, the strong model objectively evaluates the conversation against the goals and re-injects improvement guidance until met (met+score double gate) or `maxRounds` reached), task-list (ordered task chain, each auto-advanced after the previous completes); work windows (global `ws window <HH:MM-HH:MM>`, cross-midnight/days supported), startAt/endAt, cross-workspace (`ws:`/`session:`/`workspaceId`/`workspacePath` targets); web console `/watchdog` + `/api/watchdog/`; consecutive eval failures terminalize to `failed`* (observable)

Install & Auto-Enable

dsh plugin is the supported way: it installs the package into the profile, and because this package declares dsh.bundle.patch (= cordis.patch.yml), it is automatically appended to dsh.profile.bundles — enabled on next boot with no manual profile edits.

From npm (published):

# install the plugin package as a dependency
yarn add dsh-cortex            # or: npm install dsh-cortex

# register it into a DSH profile and boot
npx @deepseek-ai/dsh plugin --profile web add dsh-cortex
npx @deepseek-ai/dsh web        # boot — the 13 cortex_* tools are live

From a local checkout (development):

cd <repo-path>                                   # the repo root
npx @deepseek-ai/dsh plugin --profile web add .  # anchors to this dir (or pass the absolute path)
npx @deepseek-ai/dsh web

Remove: npx @deepseek-ai/dsh plugin --profile web remove dsh-cortex.

The shipped bundle patch inserts the cortex row with uiEnabled: true, uiPort: 8788 (console listens on loopback only). stateDir and modelsFile use built-in defaults ($DSH_HOME/storages/cortex and the bundled config/models.yaml) — anything can be overridden by an - id: cortex config row in your profile's own cordis.patch.yml (it applies after bundle layers).

Typical Flow (agent view)

cortex_start → (cortex_strategy_match) → cortex_decompose T0 → cortex_execute [node_ids | recursive]
→ cortex_review (batch accept) → cortex_recover (on failures) → cortex_report (KPI) → cortex_policy (settle)

Operations & Commands

1. Agent tools (13 — the supervisor calls them autonomously)

Tool	Purpose	Key parameters
`cortex_start`	Create a task (profile → fingerprint → strategy match → budget pool)	`goal`, `profile{task_type, domain, complexity, risk, quality_requirement, budget_limit, input_tokens?, expected_output_tokens?, latency_requirement_ms?, language?, capability?}`, `input`
`cortex_strategy_match`	Policy-center lookup (3-tier reuse verdict)	`profile`
`cortex_decompose`	Split a parent node (safety limits / DAG / replace; large-text split: `children[].split {count?, by?}` auto-splits into N parallel shard executors + 1 aggregator)	`task_id`, `parent_id?`, `replace?`, `children[{goal, task_type, capability, output_schema, quality_target?, tool_allow?, depends_on?, type?, split?}]`
`cortex_execute`	Route & execute (Gate0/1[/2], cache, accounting)	`task_id`, `node_id?`/`node_ids?`, `recursive?`, `mode?` (costFirst
`cortex_review`	Gate3 batch acceptance (decisions carry quality/self-check snapshots; Gate3.5 red-team)	`task_id`, `decisions[{node_id, status, reason_code?, missing_items?, next_action, recommended_depth?, confidence?}]`, `auto?`, `red_team?`
`cortex_recover`	Recovery decision tree (retry/switch/upgrade re-run in-engine)	`task_id`, `node_id?`
`cortex_policy`	Settle a policy from a successful task	`task_id`, `note?`
`cortex_report`	KPI report (per task or global)	`task_id?`
`cortex_models`	Model registry / profiles / routing scores / drift / price override	`task_type?`, `capability?`, `model?`, `enabled?`, `refresh?`, `reset?`, `price?`, `clear_price?`
`cortex_evaluate$	\text{Model} \text{evaluation} (7-\text{dim} \times \text{capability} \text{profiles}, \text{value}/\text{Pareto})	$cases?`,` use_saved?`,` models?`,` task_type?`,` quality_target?`
`cortex_eval_cases`	Eval-case library (page + conversation shared)	`action?` (list
`cortex_capabilities`	Capability catalog (append-only)	`action?` (list
`cortex_l0`	L0 deterministic tool library (list/show/add/remove/generate/run)	`action?`, `keywords`, `allowed?`

2. Slash commands (`/cortex <subcommand>`)

Command	What it does
`/cortex status`	Engine state: tasks (running), nodes, attempts, policies, models, discipline status
`/cortex task <id>`	Task tree (profile/strategy/budget/nodes)
`/cortex trace <id>`	Execution trail (attempts + decisions)
`/cortex kpi [id]`	KPI report (global or per task)
`/cortex models [enable\|disable <id>\|refresh\|reset\|set-price <id> <in> <out> [cached]\|clear-price <id>]`	Registry listing; `reset` clears evaluation feedback; enable/disable toggles; refresh re-discovers dynamic models; `set-price`/`clear-price` set/clear a model's effective price (persisted, 3-tier priority)
`/cortex policies [disable	enable ]`
`/cortex model-pref [set <id> [reason]\|clear\|show]`	Set/clear the preferred executor model with preference content (`reason` = which tasks it suits): routing adopts it when the task scene is covered by the content (>=50%) + active + quality <0.9 (user preference is authoritative — capability tags are advisory only, `prefCapabilityMismatch` audited; failed executions auto-upgrade via the recovery engine); persisted; same as page ★
`/cortex l0`	L0 deterministic tool library (list/show/remove/generate/run)
`/cortex rate [refresh]`	USD→CNY rate (refresh forces fetch)
`/cortex reset`	Delete state files + clear in-memory store + reload static registry + immediately re-discover models from DSH (no host restart needed); newly discovered third-party models are dedup-merged into `config/models.yaml` (append-only, comments preserved) so they join the static registry permanently
`/cortex ui start	stop
`/cortex watchdog <add task\|add goal\|list\|rm\|pause\|resume\|run\|goal\|ws\|ui>`	Watchdog: cross-workspace scheduled task injection (at/after/every, ≥300s) + goal-loop auto evaluation & remediation until reached (startAt/endAt optional; global work window 09:00–18:00; `ws window` to set/clear; consecutive eval failures terminalize to "failed" for visibility)
`/cortex routing [suggest\|show\|apply\|rollback]`	Route-weight learning: sample-driven weights suggestion (suggest), show current, one-click apply, rollback — automation driven by per-node routing samples + drift
`/cortex budget`	Budget rebalancing: shared-pool balance + ledger (dynamic reclaim/re-distribution)
`/cortex calibration`	Evaluator-calibration chain (historical contradictory samples across review vs self-check vs terminal state)
`/cortex help`	Command usage

3. Console REST (127.0.0.1:8788)

GET  /api/overview · /api/timeline?hours=all|24|168 · /api/rate · /api/models
     /api/matrix · /api/evaluations · /api/eval-cases · /api/capabilities
     /api/evaluate/status · /api/task/<id> · /api/events (SSE live)
POST /api/models/<id>            {enabled?: boolean, preferred?: boolean, reason?: string}   // reason = preference content (required on the page)
     /api/matrix/reset           (clear evaluation feedback)
     /api/eval-cases             {action: add|remove|save, ...}
     /api/capabilities           {name, description?}
     /api/evaluate               {models?, cases?, use_saved?} (real background run)
     /api/watchdog               {action: add-task|add-goal|pause|resume|remove|run|window-set|window-clear|ws-add, ...} (whitelisted)
GET  /watchdog · /api/watchdog/rules|runs?limit=50|targets|windows   (watchdog page, uses /api/events SSE)

4. Runtime lifecycle

Mount: apply → registry load (static YAML + persisted overrides + model states) → heartbeat stateDir/mounted.json → optional console → usage-settle timer (60s, zombie sweep every 5 ticks) → dynamic model discovery.
Per turn: orchestration-discipline context is injected per conversation session (suppressed only when that session has a flow active within 15 min); other sessions' leftover flows never suppress.
Unmount / restart: effect dispose stops the console (active SSE connections destroyed — no hang) and clears timers; everything durable lives in the state dir (tasks/nodes/policies/profiles/model states/rate) and rebuilds on restart (last-write-wins).
Self-healing: zombie tasks (running, 4h silent) are finalized automatically; evaluation feedback and drift baselines survive restarts; /cortex reset wipes state files for a fresh start (instant: store cleared, static registry reloaded, then DSH re-discovered); dynamic model discovery also refreshes automatically once a day (modelRefreshHours, default 24, 0=off).

Model Allocation (sub-agent pipeline)

Sensitivity tiers (driven by qualitySensitiveOf): S (complex planning / arbitration / final review / correction — needs strong reasoning), M (planning / numeric-statistics / key evaluation / visual detail — weak models may repeat & vote), L (regular execution). Decision records carry sensitivity.
Strongest fallback chain (strongestOf): configured strongestModel (must be active) → else derived from live profiles (tier → defaultQuality → measured quality → exec success rate). If no strongest model is configured, the current best available model IS the strongest — S-tier roles never stall; the mapping shifts dynamically as profiles update.
Sub-agents per node: planAgent (cortex_decompose auto:true — strongest model generates children), reviewAgent (cortex_review auto:true — threshold auto-accept for regular nodes + review sub-agent for anomalies/high-sensitivity), fixAgent (config.fixAgentAuto — on escalation/upgrade the strongest model may deliver a fix or the corrected answer directly).
Weak-model voting (repeat_vote, M/S tiers): the same model executes N rounds (2-5) and results are merged by agreement (pass/evaluate/arbitrate) — reliability of a strong model at ~1/100 cost.
Deterministic statistics (V8.0 §5.1 L0 tier): statistics/counting nodes become "extract (model, vote/early-stop) → majority-merged records → engine computation" — the model only extracts records (EXTRACT_CONTRACT); counts/aggregations are computed by lib/engine/compute.js (zero cost, replayable), eliminating model-count drift; type:"compute" = L0 engine node (decompose accepts compute:{ops:[...]}, inferred from goal by default) — results carry quality_source:"engine" (E tier, auto-accepted without supervisor review).
Pure-text statistics direct compute: word frequency / char / word / line counts — text is the data: the engine tokenizes and counts the source text directly (zero model calls, exact to each token, reproducible); entity-level stats still go extract+engine.
V8.0 metrics: cortex_report adds node autonomy rate (autonomyRate), S-tier cost share (cost.sShare) and voting/engine stats (voting: repeat attempts/cost, engine computes, estimated saving); UI report cards updated.
L0 capability library: AI-generated deterministic tools accumulate at runtime (stateDir/l0_capabilities.json, versioned) — execution checks first (id/name/keyword match → zero-cost engine run; explicit l0:"id" missing + allowed → strongest model generates code → sandbox verification (vm zero-trust / forbidden tokens / timeout / double-run determinism / sample+Gate0) → stored → executed; unverified code never runs); auto-disable after 3 consecutive failures; cortex_l0, /cortex l0, "L0 tools" UI tab.
Auto mode: autoMode(task) — quality_requirement≥0.9 ∥ complexity≥0.8 ∥ risk≥0.8 → qualityFirst automatically (no manual mode).
§7.1 KPIs: convergence rate / avg tree depth / cache hit rate (task-level + global, deterministic definitions); attempts now carry fromCache/cacheLevel (cache hits observable); no duplicate cache rows on hit (same-key idempotent).
§4.13 learning chain ⑤ + §7.1 global model usage: distillation dataset stateDir/distill.jsonl — L3 (strongest tier) high-quality outputs (execution / multi-vote consensus / S-delegated plan·review·fix·L0 generation) auto-collected when quality ≥ distillMinQuality (0.9), capped at 500 rows, observable via report distill{rows,models,lastAt}; globalKpi models gains tierUsage/upgradeRate/fallbackRate.
Domain-dimension profiles (§4.12 model×domain×task-type): domain_profiles[domain][taskType/cap] — EMA write-back from evaluation/Gate2 (same alpha, cross-restart deep-merge persistence); routing fallback chain domain → type-level → 7-dim mean → default (predictQuality(model, node, env.domain)); cortex_evaluate adds domain param (single-domain case set auto-inherits); cortex_models displays domain profiles.

Model Profiles (how they update)

Evaluation write-back (cortex_evaluate / one-click UI run): per task-type quality EMA + success rate + cost/latency + 7-dim capability profiles (dims_profile), persisted in profile_overrides.json (drift baselines anchored at first measurement).
Execution feedback: every real node execution (non-cache) updates deterministic counters — attempts/complete/failed + failure-type & task-type distributions, persisted in the same file; when samples >= 3 (EXEC_PROFILE_MIN_SAMPLES), routing uses the measured success rate instead of the static one (the more it is used, the truer the profile, the better the routing). Quality scores still come from evaluations only (Gate1 self-checks never pollute profiles).
POST /api/matrix/reset or /cortex models reset restores static baselines (evaluation + execution feedback).
Measured-only scores: the capability matrix shows a value only for verified capabilities (evaluation/execution write-back keys, tracked in measured); static declarations and defaultQuality never masquerade as verified scores — unverified cells display "no measured data".

Sub-Model Capabilities & Media Support

Executors (routed sub-models) are self-contained worker agents. What they can and cannot do:

Capability dimensions include media tags: image_analysis (declared by vision-capable models — image tasks filter by it) and audio_video_analysis (catalog dimension; no model declares it yet — matrix stays empty until a media-capable model/MCP is configured). Decompose image subtasks with capability: ['image_analysis'] (not document_analysis). Route-time closure: for nodes with image input, routing merges image_analysis into the need and applies an AND-hard-filter for media-exclusive capabilities — text-only models (even strong tiers) are excluded at routing (executor pre-check stays as the second line; force_models/preferred paths remain authoritative and call the user-chosen model).

Dimension	Supported
Input	text, files (read/glob/grep), images (read_image / MCP browser), URLs (web_search/web_fetch, read-only) — also base64 images (`data:image/...;base64` or `{image_base64, mime}`) auto-decoded to a real file for the executor
Tools (agent mode)	`read` `glob` `grep` `read_image` + `web_search`/`web_fetch` (read-only, default; override via `executorExtraTools`) + *`mcp__`** (any configured MCP server — e.g. Playwright screenshots) + `skill` + business-plugin prefixes (default `tssdp_`)
Policy	`mediaPolicy`: `auto` (default — media input forces agent mode + contract), `reject` (explicit fast-fail with diagnosis), `pass_through` (hand to MCP)
Output	structured JSON + optional top-level `artifacts` [{path, mime}] — engine validates existence & magic bytes (png/jpeg/webp/gif/mp4/mp3/wav/pdf); invalid → node `failed` with audit trail
By default denied	write/edit/shell (`pwsh`/`bash`/`run_code`) and re-orchestration (`subagent`/`workflow`/`cortex_*`/…) — an explicit per-node `tool_allow` swaps in a custom allow list for artifact-producing tasks
Honest limits	DSH provider modalities are text + image only — audio/video understanding requires a media-capable MCP/multimodal route; otherwise the executor returns the `{"unsupported": true, "reason": ...}` contract (never invents content), or `mediaPolicy: reject` fails fast with a clear diagnosis

Example flows: image analysis → cortex_execute on a node with image references (agent mode + read_image/MCP); screenshot/poster generation → Playwright MCP + artifacts declaration (magic-byte validated); audio/video processing → only with a media MCP configured (else explicit diagnosis, classified INPUT_FAILURE).

Preferred-model fallback: when a preferred executor (e.g. a text-only L1 model) is handed an image task, it runs first (attempt is recorded). The executor now pre-checks modality deterministically (modelAcceptsImage via host llm.resolveModelInfo): a model that does not declare image input fails fast with model lacks image modality (no generation — a text model must never be allowed to fabricate read-image evidence; observed in production: invented 800x600 / white-pixel result for a 1x1 red PNG). The recovery engine detects the gap and switches directly to a vision-capable candidate — no unnecessary decompose/upgrade round-trip.

Host prerequisite for vision: the vision model must declare inputModalities: [text, image] in the host's llm-deepseek.models entry (e.g. ~/.dsh/settings.yaml) — otherwise read_image refuses with "model does not declare image input" (the catalog default for undeclared entries is [text]).

Cost Model

Executor models (sub-models): exact TokenUsage from subagent/chat sessions (billed = input + cache read/write), priced from the registry — true token accounting.
Supervisor (primary model): real session-model usage via DSH sessionProjections/tokenMeter (the tokenUsage projection returns {totals:{uncachedInputTokens, outputTokens, cacheReadTokens, cacheWriteTokens}}; Cortex normalizes both nested and legacy flat shapes, so outputTokens is captured for accurate A-cost — fallback estimate only when no projection exists), plus a "vs direct primary-model run" baseline saving in cortex_report (85% cache-hit discounted).
Single ledger (usage.jsonl): task cost / global KPI / timeline all derive from usage rows (multi-vote, sub-agent, Gate2 and supervisor calls included); session-level conversation ledger dedupes detail rows across buckets — including across restarts via the cumulative anchor (no double-count).
Budget pools cover every cost: supervisor calls map to pools by kind (start/replan→plan, review/fix/evaluate→quality, recover/escalate→emergency), aggregate nodes→exec; the documented 90% high-cost lock is live (usage ≥90% → new L2/L3 excluded in routing, plus an exec-spend guard).
Peak pricing unified: models.yaml peak windows (×2) apply to A and sub-models alike (peakPricing config, default on; Sundays/weekends off-peak).
UI-triggered evaluations run without a supervisor (chat path) and are not counted as supervisor calls.
Effective price override (cortex_models.price/clear_price, UI, or /cortex models set-price|clear-price): a user-set unit price overrides the default, while a model/provider authoritative price wins over both; persisted (survives restarts) and used for routing cost scoring.

Development

npm test                 # node --test (engine/service/ui/watchdog unit + smoke) — 292 tests
node tools/verify-post-restart.mjs   # post-restart self-check (7 checks, incl. JSONL integrity + policy refs)
node tools/smoke-ui.mjs              # UI script syntax + SSE + model toggle

Versioning & releases: the version is read from package.json; bumping it (e.g. npm version patch) and pushing to main — after tests pass — auto-creates an annotated tag v<version>, generates a GitHub Release from the git log, and publishes to npm. Pushing the tag is not re-CI'd (no duplicate runs).

Repository Layout

app-pkg/index.js   entry forwarder (rollout cache-bypass, same pattern as dsh-tssdp)
lib/index.js       plugin entry (name/inject/Config/apply + lifecycle)
lib/service.js     CortexService facade (high-level operations behind the tools)
lib/tools.js       13 agent tools
lib/cli.js         /cortex command group
lib/ui-server.js   local web console (127.0.0.1:8788)
lib/watchdog.js    cross-workspace scheduled task injection + goal-loop remediation
lib/watchdog-cli.js /cortex watchdog command handlers
lib/engine/*.js    pure engines (fingerprint/router/budget/quality/recovery/store/executor/kpi/granularity/compute/l0/weight-learn/calibration/peak/state-reset)
config/models.yaml model registry (overridable)
test/              node --test suites
tools/             verification/dev scripts

Feedback & Issues

Found a bug, a misleading description, or a feature you'd like to see? Open an issue on GitHub:

Issue tracker: https://github.com/iguowz/dsh-cortex/issues
Please include: DSH version, Cortex version (from /cortex status), the operation you ran, and the full error/trace output.

Questions, suggestions and feature requests are welcome too.

License

MIT