minima-cli 0.12.4__tar.gz → 0.13.0__tar.gz

{minima_cli-0.12.4 → minima_cli-0.13.0}/.env.example

@@ -35,8 +35,9 @@ MINIMA_RECALL_RANK_BY=balanced
 MINIMA_RECALL_MAX_AGE_DAYS=0
 # Mubit search budget tier: low | mid | high.
 MINIMA_RECALL_BUDGET=mid
-# Log per-evidence score breakdowns (diagnostic; keep off in prod).
-MINIMA_RECALL_EXPLAIN=false
+# Fraction of recalls (0..1) that request + log Mubit's per-evidence fusion-score
+# breakdown (`recall_explain` log lines). 0 = off; try 0.05 when debugging ranking.
+MINIMA_RECALL_EXPLAIN_SAMPLE=0
 
 # --- Recommender tuning ---
 MINIMA_TAU_MIN=0.55

{minima_cli-0.12.4 → minima_cli-0.13.0}/CHANGELOG.md

@@ -4,6 +4,60 @@ All notable changes to Minima are documented here. The format follows
 [Keep a Changelog](https://keepachangelog.com/), and the project aims to follow
 [Semantic Versioning](https://semver.org/).
 
+## [0.13.0] - 2026-07-21
+
+The Mubit reinforcement train: mubit-sdk 0.13.0 adopted end to end, every free
+signal on the memory wire put to work, and the ground-truth spine's per-step gate
+verdicts start feeding Mubit as process rewards (PRs #192–#197).
+
+### Added
+- **Per-step process rewards** (server + harness, #195): the harness's red→green
+  gate verdicts now ride `POST /v1/feedback` as `step_outcomes[]` and relay to
+  Mubit's `record_step_outcome` — finer-grained reflection attribution and the raw
+  material for server-side workflow induction (procedural memory). Feedback truth
+  applies per step: deterministic/user verdicts only (judge gates excluded),
+  verified/failed only, relayed even on unlabeled turns (steps carry their own
+  provenance), capped at 32, duplicate-feedback check doubles as replay dedup.
+  Response gains `step_outcomes_recorded`.
+- **`POST /v1/diagnose` + `GET /v1/memory/health`** (server, #196): Mubit
+  introspection relays — failure lessons matching an error, and a per-namespace
+  memory-hygiene report (stale entries, contradictions, low-confidence counts,
+  promotion candidates). Both degrade to `memory_unavailable` instead of 500ing.
+  Full client parity (Python sync/async, TS SDK, TUI).
+- **Recovery-ladder memory brief** (TUI, #196): the replan rung asks `/v1/diagnose`
+  for "here's how this failed before" and injects matched lessons into the retry
+  preamble, next to the E2 local diagnostics. Fail-open.
+- **Drift + retrieval signals surfaced** (server, #193): Mubit DriftMonitor flags
+  riding every recall now emit `memory_drift:repeated` / `memory_drift:stagnant`
+  diagnostics, and the server's recall confidence feeds escalation as
+  `low_recall_confidence` (only when reported — absence is not a signal).
+- **Sampled retrieval observability** (server, #194): `MINIMA_RECALL_EXPLAIN_SAMPLE`
+  requests Mubit's per-evidence fusion-score breakdown on a sampled fraction of
+  recalls and logs it as `recall_explain` — the "why did memory rank this" tool.
+
+### Fixed
+- **Dereference now shares the recall budget** (server, #193): it was the one read
+  path with no timeout budget — a hung Mubit could stall a recommend for the full
+  30s client timeout.
+- **Supersession folds into staleness** (server, #193): a `superseded_by` entry is
+  treated as stale even when the `is_stale` flag lags the supersession write.
+
+### Changed
+- **mubit-sdk ≥ 0.13.0** (#192): the v0.12.3 keyed-lookup transport shim is deleted
+  in favor of the now-public `Client.lookup`; the TUI's `@mubit-ai/sdk` moves to
+  `^0.13.0` (transport retries now real). All remaining private `_control` handle
+  uses replaced with the typed surface (#194) — recall knobs and `record_outcome`
+  idempotency are first-class SDK parameters now.
+- **Bi-temporal outcome writes** (server, #194): `remember_outcome` passes
+  `occurrence_time` (observation time, not async-ingest landing time). Seeds are
+  deliberately NOT backdated — the seed-vs-live weighting design stands.
+- **Harness memory lane partition** (TUI, #197): harness trace writes stamp
+  `lane=harness`, keeping them explicitly partitioned from the server's typed
+  outcome records. Recall stays unfiltered until unlaned data ages out.
+- Dead `MubitMemory.get_context` surface removed (#193); recall-precision knobs
+  (`entry_types`, `rank_by`, `budget`, `max_age_days`, `explain_sample`) documented
+  in `docs/configuration.md` (#194).
+
 ## [0.12.4] - 2026-07-21
 
 ### Fixed

{minima_cli-0.12.4 → minima_cli-0.13.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: minima-cli
-Version: 0.12.4
+Version: 0.13.0
 Summary: Minima CLI: cost-aware LLM model routing — recommend cheaper models, backed by Mubit memory.
 Project-URL: Homepage, https://docs.minima.sh
 Project-URL: Documentation, https://docs.minima.sh
@@ -21,7 +21,7 @@ Classifier: Topic :: Software Development :: Libraries :: Python Modules
 Requires-Python: >=3.11
 Requires-Dist: anyio>=4.4
 Requires-Dist: httpx>=0.27
-Requires-Dist: mubit-sdk>=0.10.0
+Requires-Dist: mubit-sdk>=0.13.0
 Requires-Dist: pydantic-settings>=2.3
 Requires-Dist: pydantic>=2.7
 Requires-Dist: structlog>=24.1

{minima_cli-0.12.4 → minima_cli-0.13.0}/client_sdk/minima_client/client.py

@@ -18,6 +18,7 @@ from tenacity import (
 from minima.schemas.capabilities import CapabilitiesResponse
 from minima.schemas.common import Constraints, OutcomeLabel, TaskInput
 from minima.schemas.feedback import FeedbackRequest, FeedbackResponse
+from minima.schemas.insight import DiagnoseRequest, DiagnoseResponse, MemoryHealthResponse
 from minima.schemas.models_catalog import ModelsResponse
 from minima.schemas.recommend import RecommendRequest, RecommendResponse
 from minima.schemas.savings import CalibrationResponse, PolicyValueResponse, SavingsResponse
@@ -273,6 +274,38 @@ class MinimaClient:
         raise_for_status(resp)
         return StrategiesResponse.model_validate(resp.json())
 
+    def diagnose(
+        self,
+        error_text: str,
+        *,
+        error_type: str | None = None,
+        limit: int = 5,
+        namespace: str | None = None,
+        user_id: str | None = None,
+    ) -> DiagnoseResponse:
+        """Failure lessons matching an error — 'here's how this failed before'."""
+        req = DiagnoseRequest(
+            error_text=error_text,
+            error_type=error_type,
+            limit=limit,
+            namespace=namespace,
+            user_id=user_id,
+        )
+        resp = self._client.post("/v1/diagnose", json=req.model_dump(exclude_none=True))
+        raise_for_status(resp)
+        return DiagnoseResponse.model_validate(resp.json())
+
+    def memory_health(
+        self, namespace: str | None = None, stale_threshold_days: int = 30
+    ) -> MemoryHealthResponse:
+        """Per-namespace memory hygiene: staleness, contradictions, promotion candidates."""
+        params: dict[str, Any] = {"stale_threshold_days": stale_threshold_days}
+        if namespace is not None:
+            params["namespace"] = namespace
+        resp = self._client.get("/v1/memory/health", params=params)
+        raise_for_status(resp)
+        return MemoryHealthResponse.model_validate(resp.json())
+
     def capabilities(self) -> CapabilitiesResponse:
         resp = self._client.get("/v1/capabilities")
         raise_for_status(resp)
@@ -432,6 +465,38 @@ class AsyncMinimaClient:
         raise_for_status(resp)
         return ModelsResponse.model_validate(resp.json())
 
+    async def diagnose(
+        self,
+        error_text: str,
+        *,
+        error_type: str | None = None,
+        limit: int = 5,
+        namespace: str | None = None,
+        user_id: str | None = None,
+    ) -> DiagnoseResponse:
+        """Failure lessons matching an error — 'here's how this failed before'."""
+        req = DiagnoseRequest(
+            error_text=error_text,
+            error_type=error_type,
+            limit=limit,
+            namespace=namespace,
+            user_id=user_id,
+        )
+        resp = await self._client.post("/v1/diagnose", json=req.model_dump(exclude_none=True))
+        raise_for_status(resp)
+        return DiagnoseResponse.model_validate(resp.json())
+
+    async def memory_health(
+        self, namespace: str | None = None, stale_threshold_days: int = 30
+    ) -> MemoryHealthResponse:
+        """Per-namespace memory hygiene: staleness, contradictions, promotion candidates."""
+        params: dict[str, Any] = {"stale_threshold_days": stale_threshold_days}
+        if namespace is not None:
+            params["namespace"] = namespace
+        resp = await self._client.get("/v1/memory/health", params=params)
+        raise_for_status(resp)
+        return MemoryHealthResponse.model_validate(resp.json())
+
     async def capabilities(self) -> CapabilitiesResponse:
         resp = await self._client.get("/v1/capabilities")
         raise_for_status(resp)

{minima_cli-0.12.4 → minima_cli-0.13.0}/docs/api-reference.md

@@ -309,6 +309,50 @@ dominant_lesson_type, dominant_scope, lesson_ids[]`.
 
 ---
 
+## `POST /v1/diagnose`
+
+Failure lessons matching an error — "here's how this failed before". The harness recovery
+ladder calls this at a verified failure so the retry is briefed by memory. Degrades like
+the recommend hot path (a Mubit outage returns an empty list + `memory_unavailable`, never
+a 500).
+
+### Request — `DiagnoseRequest`
+
+| Field | Type | Default | Notes |
+|-------|------|---------|-------|
+| `error_text` | string (required) | — | The error/failure output to match. |
+| `error_type` | string | — | Optional classifier hint. |
+| `limit` | int `1–25` | `5` | |
+| `namespace` | string | — | Resolves to lane `minima:<namespace>`. |
+| `user_id` | string | — | |
+
+### Response — `DiagnoseResponse`
+
+`{ namespace, lane, failure_lessons: FailureLesson[], summary, total_failure_lessons,
+warnings[] }`, where each `FailureLesson` has `lesson_id, content, lesson_type,
+importance, confidence`.
+
+---
+
+## `GET /v1/memory/health`
+
+Per-namespace memory hygiene: entry counts, stale entries, contradictions, low-confidence
+counts, promotion candidates. Same graceful degradation as `/v1/diagnose`.
+
+### Query parameters
+
+| Param | Type | Default | Notes |
+|-------|------|---------|-------|
+| `namespace` | string | — | Resolves to lane `minima:<namespace>`. |
+| `stale_threshold_days` | int `1–365` | `30` | |
+
+### Response — `MemoryHealthResponse`
+
+`{ namespace, lane, entry_counts: {type: count}, stale_entries, contradictions,
+low_confidence_count, promotion_candidates, section_health, warnings[] }`.
+
+---
+
 ## `GET /v1/health`
 
 Always returns `200`; reports degraded state in the body. Never requires auth (in

{minima_cli-0.12.4 → minima_cli-0.13.0}/docs/configuration.md

@@ -21,6 +21,11 @@ single-tenant mode). A complete annotated template ships as
 | `MINIMA_MEMORY_RECALL_TIMEOUT_MS` | `2500` | Hard recall timeout; on breach, prior-only. Latency is embedder-bound (~100–300ms GPU, ~1.5s local CPU). |
 | `MINIMA_MEMORY_RECALL_LIMIT` | `25` | Max neighbors recalled per request. |
 | `MINIMA_RECALL_MODE` | `agent_routed` | `agent_routed` \| `direct_bypass` (faster but requires `enable_direct_search=true` on the Mubit instance — off by default on hosted Mubit). |
+| `MINIMA_RECALL_ENTRY_TYPES` | `fact,observation` | Server-side entry-type filter (seeds land as `fact`, feedback as `observation`). Empty = no filter. |
+| `MINIMA_RECALL_RANK_BY` | `balanced` | Mubit fusion strategy: `relevance` \| `freshness` \| `balanced` (empty = server default). |
+| `MINIMA_RECALL_BUDGET` | `mid` | Mubit search budget tier: `low` (<500ms) \| `mid` \| `high` (empty = server default). |
+| `MINIMA_RECALL_MAX_AGE_DAYS` | `0` | Hard recency window on recall (`min_timestamp`); `0` = no window. |
+| `MINIMA_RECALL_EXPLAIN_SAMPLE` | `0` | Fraction of recalls (0..1) that request + log Mubit's per-evidence fusion-score breakdown (`recall_explain` log lines). |
 | `MINIMA_LANE_PREFIX` | `minima` | Lane prefix; lane = `<prefix>:<namespace or "default">`. |
 | `MINIMA_SEED_LANE` | `minima:default` | Default lane for `minima-seed`. |
 

{minima_cli-0.12.4 → minima_cli-0.13.0}/packages/sdk/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@mubit-ai/minima-sdk",
-  "version": "0.12.4",
+  "version": "0.13.0",
   "description": "TypeScript SDK for the Minima recommender service (api.minima.sh) — typed /v1/* client for the recommend → run → judge → feedback loop.",
   "type": "module",
   "license": "FSL-1.1-Apache-2.0",

{minima_cli-0.12.4 → minima_cli-0.13.0}/packages/sdk/src/client.ts

@@ -12,8 +12,11 @@ import type {
   CalibrationResponse,
   CapabilitiesResponse,
   Constraints,
+  DiagnoseRequest,
+  DiagnoseResponse,
   FeedbackRequest,
   FeedbackResponse,
+  MemoryHealthResponse,
   ModelsResponse,
   OutcomeLabel,
   PolicyValueResponse,
@@ -277,6 +280,18 @@ export class MinimaClient {
     return this.get<StrategiesResponse>("/v1/strategies", dropUndefined(opts));
   }
 
+  /** Failure lessons matching an error — "here's how this failed before". */
+  diagnose(req: DiagnoseRequest): Promise<DiagnoseResponse> {
+    return this.post<DiagnoseResponse>("/v1/diagnose", req);
+  }
+
+  /** Per-namespace memory hygiene: staleness, contradictions, promotion candidates. */
+  memoryHealth(
+    opts: { namespace?: string; stale_threshold_days?: number } = {},
+  ): Promise<MemoryHealthResponse> {
+    return this.get<MemoryHealthResponse>("/v1/memory/health", dropUndefined(opts));
+  }
+
   models(
     opts: {
       provider?: string;

{minima_cli-0.12.4/packages/tui/src/minima → minima_cli-0.13.0/packages/sdk/src}/schemas.ts

@@ -138,6 +138,22 @@ export interface RecommendResponse {
 // Feedback — feedback.py
 // ---------------------------------------------------------------------------
 
+/**
+ * One plan step's objective verdict, relayed to Mubit as a process reward.
+ * Send only deterministic verdicts (red->green gate results) — never model
+ * self-assessment.
+ */
+export interface StepOutcome {
+  step_id: string;
+  step_name?: string;
+  outcome: OutcomeLabel;
+  /** Reinforcement signal in [-1, 1]; derived from outcome when omitted. */
+  signal?: number;
+  rationale?: string;
+  /** Corrective directive for future attempts at this step. */
+  directive_hint?: string;
+}
+
 export interface FeedbackRequest {
   recommendation_id: string;
   chosen_model_id: string;
@@ -165,6 +181,11 @@ export interface FeedbackRequest {
   chosen_effort?: string;
   notes?: string;
   idempotency_key?: string;
+  /**
+   * Per-step objective verdicts (gate results) relayed as process rewards.
+   * Independent of the turn label: recorded even when the turn is unlabeled.
+   */
+  step_outcomes?: StepOutcome[];
 }
 
 export interface FeedbackResponse {
@@ -174,6 +195,7 @@ export interface FeedbackResponse {
   updated_confidence?: number;
   reflection_triggered?: boolean;
   lesson_promoted?: boolean;
+  step_outcomes_recorded?: number;
   warnings?: string[];
 }
 
@@ -279,6 +301,47 @@ export interface Strategy {
   lesson_ids?: string[];
 }
 
+// ---------------------------------------------------------------------------
+// Memory insight — insight.py
+// ---------------------------------------------------------------------------
+
+export interface DiagnoseRequest {
+  error_text: string;
+  error_type?: string;
+  limit?: number;
+  namespace?: string;
+  user_id?: string;
+}
+
+export interface FailureLesson {
+  lesson_id?: string;
+  content?: string;
+  lesson_type?: string;
+  importance?: string;
+  confidence?: number;
+}
+
+export interface DiagnoseResponse {
+  namespace?: string;
+  lane: string;
+  failure_lessons?: FailureLesson[];
+  summary?: string;
+  total_failure_lessons?: number;
+  warnings?: string[];
+}
+
+export interface MemoryHealthResponse {
+  namespace?: string;
+  lane: string;
+  entry_counts?: Record<string, number>;
+  stale_entries?: number;
+  contradictions?: number;
+  low_confidence_count?: number;
+  promotion_candidates?: number;
+  section_health?: Record<string, unknown>;
+  warnings?: string[];
+}
+
 // ---------------------------------------------------------------------------
 // Capabilities — capabilities.py
 // ---------------------------------------------------------------------------

{minima_cli-0.12.4 → minima_cli-0.13.0}/packages/tui/bun.lock

@@ -7,7 +7,7 @@
       "dependencies": {
         "@anthropic-ai/sdk": "^0.40.0",
         "@google/genai": "^0.3.0",
-        "@mubit-ai/sdk": "^0.10.0",
+        "@mubit-ai/sdk": "^0.13.0",
         "ink": "^5.0.0",
         "react": "^18.3.0",
         "string-width": "^7.2.0",
@@ -51,7 +51,7 @@
 
     "@js-sdsl/ordered-map": ["@js-sdsl/ordered-map@4.4.2", "", {}, "sha512-iUKgm52T8HOE/makSxjqoWhe95ZJA1/G1sYsGev2JDKUSS14KAgg1LHb+Ba+IPow0xflbnSkOsZcO08C7w1gYw=="],
 
-    "@mubit-ai/sdk": ["@mubit-ai/sdk@0.10.0", "", { "dependencies": { "@grpc/grpc-js": "^1.14.1", "@grpc/proto-loader": "^0.8.0" } }, "sha512-DxVm0Slv6ezMQaSC8+gMcOW65+otUs8GC/nPI9wThzdf6IonfR9LzJfJY9gm7nVlvv5oVI3hOJk/2G36I201BA=="],
+    "@mubit-ai/sdk": ["@mubit-ai/sdk@0.13.0", "", { "dependencies": { "@grpc/grpc-js": "^1.14.1", "@grpc/proto-loader": "^0.8.0" } }, "sha512-lKh/k6Q94gMgJAEkl2wqFagN5ml6KVOyZwruCK1JVildM/+9OzReIigvY2IjlJTyY3CgX+VSEiMD/AkhYblrlA=="],
 
     "@protobufjs/aspromise": ["@protobufjs/aspromise@1.1.2", "", {}, "sha512-j+gKExEuLmKwvz3OgROXtrJ2UG2x8Ch2YZUxahh+s1F2HZ+wAceUNLkvy6zKCPVRkU++ZWQrdxsUeQXmcg4uoQ=="],
 

{minima_cli-0.12.4 → minima_cli-0.13.0}/packages/tui/package.json

@@ -1,6 +1,6 @@
 {
   "name": "minima-tui",
-  "version": "0.12.4",
+  "version": "0.13.0",
   "description": "Minima harness + TUI — the `minima` CLI (TypeScript/Ink; successor to the removed Python harness). Talks to the Python FastAPI recommender service over the /v1/* contract.",
   "type": "module",
   "license": "FSL-1.1-Apache-2.0",
@@ -30,7 +30,7 @@
   "dependencies": {
     "@anthropic-ai/sdk": "^0.40.0",
     "@google/genai": "^0.3.0",
-    "@mubit-ai/sdk": "^0.10.0",
+    "@mubit-ai/sdk": "^0.13.0",
     "ink": "^5.0.0",
     "react": "^18.3.0",
     "string-width": "^7.2.0"

{minima_cli-0.12.4 → minima_cli-0.13.0}/packages/tui/src/minima/client.ts

@@ -12,8 +12,11 @@ import type {
   CalibrationResponse,
   CapabilitiesResponse,
   Constraints,
+  DiagnoseRequest,
+  DiagnoseResponse,
   FeedbackRequest,
   FeedbackResponse,
+  MemoryHealthResponse,
   ModelsResponse,
   OutcomeLabel,
   RecommendRequest,
@@ -176,6 +179,17 @@ export class MinimaClient {
     return this.get<StrategiesResponse>("/v1/strategies", dropUndefined(opts));
   }
 
+  /** Failure lessons matching an error — the recovery ladder's "how this failed before" brief. */
+  diagnose(req: DiagnoseRequest): Promise<DiagnoseResponse> {
+    return this.post<DiagnoseResponse>("/v1/diagnose", req);
+  }
+
+  memoryHealth(
+    opts: { namespace?: string; stale_threshold_days?: number } = {},
+  ): Promise<MemoryHealthResponse> {
+    return this.get<MemoryHealthResponse>("/v1/memory/health", dropUndefined(opts));
+  }
+
   models(
     opts: {
       provider?: string;

{minima_cli-0.12.4 → minima_cli-0.13.0}/packages/tui/src/minima/index.ts

@@ -30,7 +30,13 @@ export {
 } from "./judge.ts";
 export type { QualityJudge } from "./judge.ts";
 export { MinimaRouter, type RoutingResult, type Ranking } from "./router.ts";
-export { MinimaAgent, gradeOutcome, type BeforeRoute, type MinimaAgentOptions } from "./runtime.ts";
+export {
+  MinimaAgent,
+  gradeOutcome,
+  stepOutcomesFromGates,
+  type BeforeRoute,
+  type MinimaAgentOptions,
+} from "./runtime.ts";
 export {
   buildMemoryProjection,
   memoryProjectionFor,

{minima_cli-0.12.4 → minima_cli-0.13.0}/packages/tui/src/minima/memory.ts

@@ -21,6 +21,17 @@ const RECALL_ENTRY_TYPES = ["lesson", "rule", "observation"];
 const MAX_SNIPPETS = 5;
 const MAX_SNIPPET_CHARS = 240;
 
+/**
+ * Write-side lane partition. Two structurally incompatible writers share one Mubit
+ * backend: the Minima SERVER writes typed OutcomeRecord JSON (intent=observation,
+ * lane=minima:<namespace>, its own run ids) and this HARNESS writes free-text traces.
+ * Stamping every harness write with its own lane keeps the two deliberately
+ * partitioned at the source, whatever run ids are in play. Recall is deliberately
+ * NOT lane-filtered yet — pre-partition entries have no lane and would vanish from
+ * recall; filter once unlaned data has aged out.
+ */
+const HARNESS_LANE = "harness";
+
 export interface OutcomeRecord {
   task: string;
   recommendationId: string;
@@ -145,6 +156,7 @@ export class MubitHarnessMemory implements HarnessMemory {
         session_id: this.sessionId,
         agent_id: this.agentId,
         idempotency_key: record.recommendationId,
+        lane: HARNESS_LANE,
       });
       // Close the loop with a real outcome score tied to the recommendation. Never
       // fabricate — a judge abstention (quality === null) records the trace but no score.

{minima_cli-0.12.4 → minima_cli-0.13.0}/packages/tui/src/minima/router.ts

@@ -18,6 +18,7 @@ import type {
   Constraints,
   FeedbackResponse,
   RankedModel,
+  StepOutcome,
   TaskInput,
   TaskType,
 } from "./schemas.ts";
@@ -250,6 +251,32 @@ export class MinimaRouter {
     });
   }
 
+  /**
+   * Failure-lesson brief for a recovery re-decision: "here's how this failed
+   * before", from Mubit's diagnose via the server. Fail-open — a memory outage
+   * or old server must never block a retry. Returns null when nothing matched.
+   */
+  async diagnoseBrief(errorText: string): Promise<string | null> {
+    const text = errorText.trim();
+    if (!text) return null;
+    try {
+      const resp = await this.client.diagnose({
+        error_text: text.slice(0, 2000),
+        limit: 3,
+        namespace: this.config.namespace ?? undefined,
+      });
+      const lessons = (resp.failure_lessons ?? [])
+        .map((l) => (l.content ?? "").trim())
+        .filter((c) => c.length > 0);
+      if (lessons.length === 0) return null;
+      return `Past failure lessons for similar errors (from memory):\n${lessons
+        .map((c) => `- ${c}`)
+        .join("\n")}`;
+    } catch {
+      return null;
+    }
+  }
+
   async feedback(opts: {
     recommendationId: string;
     chosenModelId: string;
@@ -276,6 +303,12 @@ export class MinimaRouter {
     chosenEffort?: string;
     /** Provenance tag, e.g. "unlabeled" for cadence-skipped/abstained turns. */
     notes?: string;
+    /**
+     * Per-step objective verdicts (deterministic/user gate results only — never the
+     * judge) relayed to the server as Mubit process rewards. Independent of the
+     * turn label: sent even when evidenceSource is "none".
+     */
+    stepOutcomes?: StepOutcome[];
   }): Promise<FeedbackResponse> {
     return await this.client.feedback({
       recommendation_id: opts.recommendationId,
@@ -293,6 +326,7 @@ export class MinimaRouter {
       judged: opts.judged,
       chosen_effort: opts.chosenEffort,
       notes: opts.notes,
+      step_outcomes: opts.stepOutcomes?.length ? opts.stepOutcomes : undefined,
     });
   }
 }

{minima_cli-0.12.4 → minima_cli-0.13.0}/packages/tui/src/minima/runtime.ts

@@ -57,6 +57,7 @@ import { knownProcedureFor } from "./memory_dream.ts";
 import { memoryProjectionFor } from "./memory_ledger.ts";
 import type { CostMeter } from "./meter.ts";
 import { MinimaRouter, type RoutingResult } from "./router.ts";
+import type { StepOutcome } from "./schemas.ts";
 import { makeStopGate } from "./stop_gate.ts";
 
 /** Inspect/override a recommendation before the model runs. Return a result to override;
@@ -98,6 +99,36 @@ export function buildFeedbackNotes(
   return parts.length > 0 ? parts.join(";") : undefined;
 }
 
+/** Map this rung's gate rows to wire step outcomes (Mubit process rewards).
+ *
+ * Feedback truth applies per step exactly as per turn: only deterministic/user
+ * verdicts count (a judge gate is model self-assessment — excluded), and only
+ * verified/failed are evidence (unrunnable/unchecked are environmental). Multiple
+ * gates on one step collapse to the LAST verdict (rows arrive in created_at order),
+ * mirroring the red→green flip semantics of groundedOutcomeFor. */
+export function stepOutcomesFromGates(
+  gates: readonly {
+    step_id: string | null;
+    kind: string | null;
+    outcome: string | null;
+    confidence: string | null;
+    verified_by: string | null;
+  }[],
+): StepOutcome[] {
+  const finals = new Map<string, StepOutcome>();
+  for (const g of gates) {
+    if (!g.step_id) continue;
+    if (g.verified_by !== "deterministic" && g.verified_by !== "user") continue;
+    if (g.outcome !== "verified" && g.outcome !== "failed") continue;
+    finals.set(g.step_id, {
+      step_id: g.step_id,
+      outcome: g.outcome === "verified" ? "success" : "failure",
+      rationale: `${g.kind ?? "gate"}/${g.verified_by}${g.confidence ? `/${g.confidence}` : ""}`,
+    });
+  }
+  return [...finals.values()].slice(0, 32);
+}
+
 export function gradeOutcome(quality: number): "success" | "partial" | "failure" {
   if (quality >= 0.8) return "success";
   if (quality >= 0.4) return "partial";
@@ -642,6 +673,10 @@ export class MinimaAgent extends Agent {
           // output so the retry reasons from evidence. Feedback DESIGN over retry count.
           const diag = await this.collectFailureDiagnostics();
           if (diag) replanPrefix += `\n\n${diag}`;
+          // Memory brief: failure lessons Mubit already learned for errors like this one
+          // (server /v1/diagnose). Fail-open; nothing matched → nothing injected.
+          const brief = await this.router.diagnoseBrief(errorText ?? diag ?? "");
+          if (brief) replanPrefix += `\n\n${brief}`;
           if (decision) writeRecoveryGate(this.recoveryGateDeps(), decision);
           this.ladderReplans += 1;
         } else {
@@ -985,6 +1020,17 @@ export class MinimaAgent extends Agent {
       //  - the legacy judged/verifiedInProduction flags ride along for old servers.
       const judged = evidenceSource === "judge";
       const verifiedInProduction = evidenceSource === "gate";
+      // Per-step process rewards: this rung's deterministic/user gate verdicts, keyed
+      // by the same rec_id identity join as the grounded outcome. Fail-open — a DB
+      // read must never sink feedback.
+      let stepOutcomes: StepOutcome[] | undefined;
+      try {
+        const gateRows = this.db?.getGatesForRec(routing.recommendationId) ?? [];
+        const collected = stepOutcomesFromGates(gateRows);
+        stepOutcomes = collected.length > 0 ? collected : undefined;
+      } catch {
+        stepOutcomes = undefined;
+      }
       const resp = await this.router.feedback({
         recommendationId: routing.recommendationId,
         chosenModelId: routing.chosenModelId,
@@ -999,6 +1045,7 @@ export class MinimaAgent extends Agent {
         judged,
         chosenEffort: this.agentState.thinkingLevel ?? undefined,
         notes: buildFeedbackNotes(deterministic, judged, recoveryRung),
+        stepOutcomes,
       });
       // Keep the Mubit-side provenance ids (previously discarded) — the work record
       // cites which memory entries this outcome reinforced.