pantheon-opencode 1.5.0-beta.9 → 1.5.1

package/src/agents/zeus.md

@@ -3,6 +3,7 @@ description: "Orquestrador central — NUNCA implementa. Roteia para especialist
 mode: primary
 reasoning_effort: medium
 permission:
+  read: allow
   edit: deny
   bash: deny
   task:
@@ -144,33 +145,11 @@ task_status(task_id="ses_xxx", wait=true)
   -> bloqueia ate completar: { state: "completed", task_result: "..." }
 ```
 
-### Pantheon Delegation Tools (plugin) — 3-tool API
+### DELEGATION_RULES (native task API)
 
-O plugin Pantheon adiciona 3 tools de delegacao gerenciada pelo BackgroundJobBoard:
-
-```
-pantheon_delegate({prompt, agent, description?, read_only?})
-  -> cria sessao filha, registra no board, retorna alias: [apo-1]
-pantheon_delegation_read({id: "apo-1"})
-  -> BLOQUEIA ate o job terminar, retorna o report markdown, marca reconciled
-  -> durante a espera, mostra o que o agente esta fazendo: seccao final "## Agent Activity"
-     (amostra as messages do child a cada ~2s; tool calls + args truncados / texto)
-pantheon_delegation_list()
-  -> lista jobs da sessao, com [unread] para jobs terminados nao lidos
-  -> jobs RUNNING mostram "last activity:" (ultima acao visivel do child)
-  -> fail-open: sem suporte a messages, formato atual inalterado
-```
-
-- **Read-only enforcement:** agentes read-only (apollo, gaia — `read_only_agents` no routing.yml) tem `edit`, `write`, `bash`, `task` NEGADOS dentro da sessao delegada (o guard de `tool.execute.before` THROW com mensagem acionavel). Use `read_only: true` no delegate para qualquer agente de investigacao.
-- **Depth-2 hard-enforced:** sessao read-only NAO cria subagentes (`task` bloqueado) — investigacao nunca vira arvore.
-- **Compactacao:** jobs em voo (running + unread terminal) sao injetados no contexto de compactacao via `experimental.session.compacting` — delegacao em andamento nao se perde ao compactar.
-
-### DELEGATION_RULES (plugin path)
-
-1. **NUNCA** polle `pantheon_delegation_list` para checar se um job terminou.
-2. Voce NAO recebe notificacao injetada no chat — politica do usuario: ZERO `task-notification` no transcript. Visibilidade de conclusao: marcador `[unread]` em `pantheon_delegation_list`, `pantheon_delegation_read({id})`, toasts TUI (pantheon-hooks) e carry-forward de compactacao.
-3. Use `pantheon_delegation_read({id})` APENAS para fan-in explicito / recuperacao de resultado sob demanda (bloqueia ate o fim).
-4. `pantheon_delegation_list` e para diagnostico, nao para polling.
+1. Todo dispatch usa `task(background=true, subagent_type=..., prompt=...)`.
+2. Recolha com `task_status(task_id=..., wait=true)` — bloqueia ate o fim, sem polling.
+3. Use `task_status(wait=true)` apenas para fan-in explicito / recuperacao sob demanda.
 
 ### Background (sempre usar)
 - **Apollo, Hermes, Aphrodite, Demeter, Hephaestus, Prometheus**
@@ -301,32 +280,11 @@ vão a idle com todos incompletos. Guards (todos no `src/pantheon/todo-enforcer.
 
 ## Wave 4 (PR #46): Empty-Result Retry + /cost + Themis Tier
 
-### Empty-Result Retry (dispatch-guard — MANUAL orchestration)
-OpenCode 1.18.x NAO permite interceptar a conclusao de `task()` via hooks,
-entao o `src/pantheon/dispatch-guard.ts` e uma lib pura usada POR VOCE na
-orquestracao (NAO esta wired no plugin):
-
-```
-import { createDispatchGuard } from '.../src/pantheon/dispatch-guard.ts'  # via code-mode ou subagente
-
-guard = createDispatchGuard({ retryOnEmpty: true, logger: { warn: console.warn } })
-
-# classificar resultado de task_status/wait:
-#   'content'      → tem texto, pronto
-#   'empty-mode1'  → SEM texto E SEM tokens (nada voltou)      → RETRY 1x
-#   'empty-mode2'  → SEM texto, MAS tokens (raciocinou, perdeu a parte de texto;
-#                    assinatura da falha themis Wave-2)         → RETRY 1x
-
-out = await guard.maybeRetry(result, async () => { ...task() de novo... })
-# out.retried=true se redisparou; CAP DURO de 1 retry — nunca 2x.
-# Se o retry voltar vazio, `out.retried=false` — NAO tente de novo: escale.
-```
-Regra: retry 1x APENAS em `empty-mode1`/`empty-mode2`. Resultado com conteudo
-nunca redispara. Apos 1 retry vazio → escalate (mesma regra do TODO Enforcer).
-
-> **Waves DEVEM usar `zeusDelegateWithRetry`** (`src/pantheon/zeus-delegate-with-retry.ts`) — encapsula `delegate → waitForTerminal → classify → retry 1x → escalate` com cap duro 1; se ainda vazio throw `ZeusEscalationError` (escale: tente outro agente, simplifique, manual).
-> Uso alto-nível: `import { zeusDelegateWithRetry } from './pantheon/zeus-delegate-with-retry.ts'` → `await zeusDelegateWithRetry({ board, client, sessionID, agent, prompt })`.
-> Uso baixo-nível: `createZeusRetryHelper().executeWithRetry(first, () => secondRead())` — classifica empty e decide retry/escalate (mesmo cap 1).
+### Empty-Result Retry (runtime — automatic)
+Empty-response detection for native `task()` subagent calls is wired at runtime
+in `task-result-guard.ts` (plugin `tool.execute.after` chain): an
+empty/whitespace-only result is converted into an explicit error so a silent
+child failure is never mistaken for success. No manual retry helper is required.
 
 ### /cost — pantheon_cost tool (WIRED no plugin)
 `pantheon_cost({ days?: number })` le o `opencode.db` READ-ONLY usando o backend

package/src/instructions/memory-protocol.instructions.md

@@ -84,7 +84,9 @@ memory_store({
     agreements: ["point1", "point2"],
     divergences: [{"issue": "...", "resolution": "..."}],
     response_rate: "X of Y",
-    themis_audit: "approved|issues"
+    themis_audit: "approved|issues",
+    precedent_used: false,
+    timestamp: "<ISO-8601>"
   },
   metadata: {
     type: "council_decision",
@@ -93,6 +95,7 @@ memory_store({
   }
 })
 ```
+`value` is JSON-serialized before storing (the MCP `memory_store.value` argument is a string).
 
 ### Read Path (Precedent Fast-Path)
 Before dispatching a new council, Zeus runs:
@@ -103,8 +106,8 @@ memory_search(question, top_k=2, namespace="council_decisions")
 Result interpretation:
 | Score | Age | Action |
 |-------|-----|--------|
-| > 0.85 | < 30 days | Return precedent as fast-path answer. Skip council dispatch entirely. |
-| > 0.85 | >= 30 days | Return with warning "Reavaliar se contexto mudou" + proceed with council |
+| > 0.85 | < 30 days | Return precedent verbatim (note "⚠️ Decisão de [data] — reavaliar se contexto mudou") as fast-path answer. Skip council dispatch entirely. |
+| > 0.85 | >= 30 days | Return with warning "Reavaliar se contexto mudou — decisão tem mais de 30 dias" + proceed with council |
 | 0.5 - 0.85 | Any | Include as context for specialists but still dispatch council |
 | < 0.5 | Any | Ignore, proceed with fresh council |
 

package/src/instructions/zeus-anti-stall.instructions.md

@@ -15,7 +15,7 @@ You MUST self-monitor for these stall conditions:
 | Symptom | Detection Rule | Recovery Action |
 |---------|---------------|-----------------|
 | Silent loop | 3+ consecutive turns with no tool call AND no visible progress | Output `[STALL_DETECTED]` and re-read your task definition. If still stuck, escalate to user with: "I appear to be stuck on [task]. Options: (1) retry with different approach, (2) delegate to specialist, (3) simplify scope." |
-| Delegation black hole | Agent dispatched but no response after 2x the timeout from routing.yml | Log the hang, cancel via `cancel_task`, dispatch to fallback agent, report to user |
+| Delegation black hole | Agent dispatched but no response after 2x the timeout from routing.yml | Log the hang, cancel via `cancel_task`, then follow the fallback chain in `## ⏱️ Timeout & Retry Enforcement` |
 | Circular delegation | Same specialist re-dispatched for same task 2+ times without progress | Break cycle: dispatch to different specialist OR escalate to user |
 | Idle after completion | All background tasks completed but no synthesis/next step for 2+ turns | Force synthesis: summarize all completed results and propose next action |
 | Context thrash | Re-reading same files repeatedly without new action | Stop re-reading. State: "Already have context on [file]. Proceeding with [action]." |
@@ -41,9 +41,7 @@ When a delegation fails (timeout, empty response, error):
 
 2. **Retry ONCE** with rephrased prompt — add: "Previous attempt failed with: [error]. Adjusted approach: [what changed]."
 
-3. **If retry also fails** → DO NOT retry a third time blindly. Instead:
-   - Dispatch to fallback agent (from routing.yml)
-   - If no fallback, escalate to user with: "Task [X] failed after retry. Options: (a) simplify, (b) different agent, (c) manual intervention."
+3. **If retry also fails** → do NOT retry blindly; follow the fallback chain and escalation protocol in `## ⏱️ Timeout & Retry Enforcement`.
 
 ## Progress Checkpoint
 

package/src/instructions/zeus-council-synthesis.instructions.md

@@ -85,10 +85,7 @@ sequenceDiagram
 ## Dispatch Sequence (9-Step Protocol)
 
 ### Step 0 — Precedent Fast-Path (Fase 1)
-Before any dispatch: `memory_search(query, top_k=2, namespace="council_decisions")`
-- Score > 0.85 AND age < 30 days → return precedent verbatim with note "⚠️ Decisão de [data] — reavaliar se contexto mudou". SKIP entire dispatch.
-- Score > 0.85 AND age > 30 days → return precedent with ⚠️ "Reavaliar se contexto mudou — decisão tem mais de 30 dias"
-- Else → proceed to Step 0b
+Run the precedent read path defined in `## Memory Protocol > Council Decisions Namespace` (`memory_search(query, top_k=2, namespace="council_decisions")`). If no precedent applies, proceed to Step 0b.
 
 ### Step 0b — Apollo Pre-Scan (Fase 2, --research flag)
 If `/pantheon --research <question>`: dispatch @apollo with 30s timeout. Inject findings as `shared_context` into ALL specialist prompts. Skip if flag absent.
@@ -190,33 +187,11 @@ task(subagent_type: "themis", prompt: "Audit this council synthesis for fidelity
 - If issues → fix each issue before delivering to user
 
 ### Step 9 — Persist & Reconcile (Fase 1)
-```
-memory_store({
-  namespace: "council_decisions",
-  key: "council:<yyyy-mm-dd>:<slug>",
-  value: JSON.stringify({
-    question, specialists, recommendation, confidence,
-    agreements, divergences, precedent_used, timestamp
-  }),
-  metadata: {type: "council_decision", specialist_count: N}
-})
-board.markReconciled("<task-id>")
-```
+Persist the decision using the write path in `## Memory Protocol > Council Decisions Namespace`, then `board.markReconciled("<task-id>")`.
 
 ## Specialist Output Format (Fase 1 — machine-parseable structured fields)
 
-Specialists MUST return these structured fields in their response:
-
-```
-## specialist_response
-**position:** <clear one-sentence position>
-**reasoning:** <2-4 sentences>
-**trade_offs:** <what's gained vs lost>
-**risks:** <what could go wrong>
-**confidence:** High | Medium | Low
-**agreement_signals:** agree: @agent1, @agent2 on [issue] | disagree: @agent3 on [issue]
-**specific_claims:** <count of specific factual claims in response>
-```
+Specialists MUST return the structured `## specialist_response` format (position, reasoning, trade_offs, risks, confidence, agreement_signals, specific_claims) defined in `## Agent Return Format > Council Specialist Response Format`.
 
 ## Domain-to-Specialist Mapping
 

package/src/instructions/zeus-timeout-retry.instructions.md

@@ -126,39 +126,14 @@ ANY NO  → full task (IMPL artifact + Themis review mandatory)
 3. **No Themis dependency** — output doesn't feed into a phase that requires review
 
 ## Subtask Return Format
-Expect a `subtask_summary` response with:
-```
-## subtask_summary
-**files_changed:** [paths]
-**summary:** What was done
-**tests:** ✅ or N/A
-**status:** complete | partial | escalated
-```
+Return the standard `## subtask_summary` defined in `## Agent Return Format` (files_changed, summary, tests, coverage, tokens, status, blockers).
 
 ## Timeout Parcial (Partial Results)
 
-Timeout parcial is ONLY for read-only, independent agents:
-- ✅ @apollo — can return partial file list ("found 7 of 12 files before timeout")
-- ✅ @gaia — can return partial literature findings
-
-- ✅ @talos — can confirm progress if hotfix times out
-- ❌ Never for implementers or reviewers — must complete or fail
-
-When dispatching with partial-OK, set expectation:
-```
-@apollo Search for auth files. Timeout parcial OK — return whatever you have.
-```
+Only agents marked ✅ in the Timeout Behavior table above may return partial results: @apollo (partial file list, e.g. "found 7 of 12 files before timeout"), @gaia (partial literature findings) and @talos (confirm progress if a hotfix times out). Never for implementers or reviewers — they must complete or fail. When dispatching with partial-OK, set the expectation: `@apollo Search for auth files. Timeout parcial OK — return whatever you have.`
 
 ---
 
 # 📊 TIMEOUT TRACKING
 
-Maintain awareness of in-flight delegations:
-
-| Agent | Timeout | Status | Partial OK? |
-|-------|---------|--------|-------------|
-| @apollo | 60s | ✅ complete | ✅ |
-| @hermes | 180s | ⏳ in progress | ❌ |
-| @themis | 120s | ⏳ in progress | ❌ |
-
-Log timeouts to `/memories/session/timeout-log.md` for later analysis.
+Track in-flight delegations against the Timeout Behavior table above. Log timeouts to `/memories/session/timeout-log.md` for later analysis.

package/src/mcp/mcp_persistence_server.py

@@ -1323,7 +1323,7 @@ def build_rehydration_blocks(checkpoint: dict) -> list[str] | None:
     """Build deterministic rehydration blocks from a checkpoint dict.
 
     Pure string templating — no LLM, no embeddings. Mirrors the
-    ``<mission_context>`` convention of delegation-compaction.ts: only a
+    ``<mission_context>`` convention of the compaction-context builder: only a
     non-done goal is re-emitted. Returns None when nothing critical remains.
     """
     if not isinstance(checkpoint, dict) or not checkpoint:

package/src/mcp/memory_mcp_server.py

@@ -6,6 +6,10 @@ for semantic memory with hybrid search (vector cosine + FTS5 BM25).
 
 Total footprint: ~50MB (sqlite-vec + fastembed) vs ~1.4GB (old chromadb).
 
+Both embedding backends are optional: if either fails to import the server
+still starts and serves FTS5-only search (``PANTHEON_MEMORY_EMBED=off``
+disables the vector pipeline explicitly).
+
 Tools:
     memory_store   — Store with automatic embedding
     memory_search  — Hybrid vector + FTS5 keyword search
@@ -30,9 +34,21 @@ from datetime import UTC, datetime
 from pathlib import Path
 from typing import Any
 
-import sqlite_vec
+# Embedding backends are optional (issue #159): a failed fastembed/sqlite_vec
+# install (network error, incompatible wheel) must not kill the server — it
+# degrades to FTS-only search instead of crashing before the MCP handshake.
+try:
+    import sqlite_vec
+except ImportError:  # pragma: no cover - degraded mode
+    sqlite_vec = None  # type: ignore[assignment]
+
 from _pantheon_paths import pantheon_home
-from fastembed import TextEmbedding
+
+try:
+    from fastembed import TextEmbedding
+except ImportError:  # pragma: no cover - degraded mode
+    TextEmbedding = None  # type: ignore[assignment]
+
 from mcp.server.fastmcp import FastMCP
 
 try:
@@ -52,6 +68,12 @@ except ImportError:  # pragma: no cover - deployed runtime may need path fix
 DB_PATH = pantheon_home() / "memory" / "memory.db"
 EMBED_CACHE = Path.home() / ".cache" / "fastembed"
 
+# PANTHEON_MEMORY_EMBED=off disables the embedding/vector pipeline so the
+# server serves FTS-only search (issue #159). Default "on" keeps the current
+# hybrid vector + FTS behavior. The flag is also forced off automatically
+# when fastembed failed to import.
+_EMBED_DISABLED = frozenset({"off", "0", "false", "no"})
+
 
 _BYTE_UNIT = 1024
 
@@ -81,11 +103,6 @@ CREATE TABLE IF NOT EXISTS memories (
 CREATE UNIQUE INDEX IF NOT EXISTS idx_memories_ns_key
     ON memories(namespace, key);
 
-CREATE VIRTUAL TABLE IF NOT EXISTS vec_memories USING vec0(
-    id INTEGER PRIMARY KEY,
-    embedding float[384]
-);
-
 CREATE VIRTUAL TABLE IF NOT EXISTS memories_fts USING fts5(
     namespace, key, value, metadata,
     content='memories', content_rowid='id',
@@ -113,6 +130,14 @@ CREATE TRIGGER IF NOT EXISTS memories_au AFTER UPDATE ON memories BEGIN
 END;
 """
 
+# Vector store requires the sqlite-vec extension; created only when available.
+VEC_SCHEMA_SQL = """
+CREATE VIRTUAL TABLE IF NOT EXISTS vec_memories USING vec0(
+    id INTEGER PRIMARY KEY,
+    embedding float[384]
+);
+"""
+
 # ── FastMCP App ───────────────────────────────────────────────────────────────
 
 mcp = FastMCP(
@@ -129,7 +154,8 @@ def _get_db() -> sqlite3.Connection:
     """Get or create the SQLite connection singleton.
 
     Creates DB directory, applies WAL/performance pragmas, registers
-    the sqlite-vec extension, and runs schema init.
+    the sqlite-vec extension, and runs schema init. When sqlite-vec is
+    unavailable the vector table is skipped — search degrades to FTS-only.
     """
     DB_PATH.parent.mkdir(parents=True, exist_ok=True)
     conn = sqlite3.connect(str(DB_PATH), check_same_thread=False)
@@ -137,9 +163,10 @@ def _get_db() -> sqlite3.Connection:
     conn.execute("PRAGMA journal_mode=WAL")
     conn.execute("PRAGMA busy_timeout=5000")
     conn.execute("PRAGMA foreign_keys=ON")
-    # Register sqlite-vec extension
-
-    sqlite_vec.load(conn)
+    # Register sqlite-vec extension (optional; enables vector search)
+    if sqlite_vec is not None:
+        sqlite_vec.load(conn)
+        conn.executescript(VEC_SCHEMA_SQL)
     conn.executescript(SCHEMA_SQL)
     if _codemap is not None:
         with contextlib.suppress(Exception):
@@ -155,13 +182,30 @@ def _get_conn() -> sqlite3.Connection:
 # ── Embedding ─────────────────────────────────────────────────────────────────
 
 
+def _embeddings_enabled() -> bool:
+    """Whether the embedding/vector pipeline is active.
+
+    Controlled by ``PANTHEON_MEMORY_EMBED`` (``on`` by default; ``off``,
+    ``0``, ``false`` or ``no`` disable it). Automatically disabled when
+    ``fastembed`` failed to import, so the server degrades to FTS-only
+    search instead of crashing at startup (issue #159).
+    """
+    if TextEmbedding is None:
+        return False
+    flag = os.environ.get("PANTHEON_MEMORY_EMBED", "on").strip().lower()
+    return flag not in _EMBED_DISABLED
+
+
 @functools.cache
-def _get_embedder() -> TextEmbedding:
+def _get_embedder() -> TextEmbedding | None:
     """Lazy-load fastembed model (auto-downloads on first call, ~30MB).
 
     Uses BAAI/bge-small-en-v1.5 (384-dim, CPU, ONNX) — no PyTorch needed.
-    Model is cached in ~/.cache/fastembed/.
+    Model is cached in ~/.cache/fastembed/. Returns None when fastembed is
+    unavailable.
     """
+    if TextEmbedding is None:
+        return None
     os.environ.setdefault("TQDM_DISABLE", "1")
     return TextEmbedding(
         model_name="BAAI/bge-small-en-v1.5",
@@ -169,9 +213,18 @@ def _get_embedder() -> TextEmbedding:
     )
 
 
-def _embed(text: str) -> list[float]:
-    """Generate a 384-dim embedding vector for the given text."""
-    return next(iter(_get_embedder().embed(text))).tolist()
+def _embed(text: str) -> list[float] | None:
+    """Generate a 384-dim embedding vector for the given text.
+
+    Returns None when embeddings are disabled (``PANTHEON_MEMORY_EMBED=off``
+    or fastembed missing) — callers fall back to FTS-only search.
+    """
+    if not _embeddings_enabled():
+        return None
+    embedder = _get_embedder()
+    if embedder is None:
+        return None
+    return next(iter(embedder.embed(text))).tolist()
 
 
 # ── Helpers ───────────────────────────────────────────────────────────────────
@@ -341,14 +394,14 @@ def memory_store(
         )
         entry_id = cur.lastrowid
 
-        # Generate embedding
+        # Generate embedding (None when disabled → FTS-only mode)
         embedding = _embed(value)
-        embedding_json = json.dumps(embedding)
-
-        db.execute(
-            "INSERT INTO vec_memories(id, embedding) VALUES (?, ?)",
-            [entry_id, embedding_json],
-        )
+        if embedding is not None and sqlite_vec is not None:
+            embedding_json = json.dumps(embedding)
+            db.execute(
+                "INSERT INTO vec_memories(id, embedding) VALUES (?, ?)",
+                [entry_id, embedding_json],
+            )
         db.commit()
     except sqlite3.IntegrityError as e:
         db.rollback()
@@ -406,8 +459,8 @@ def memory_search(  # noqa: C901, PLR0912
     vec_ids: list[int] = []
     fts_ids: list[int] = []
 
-    # 2. Vector search
-    if query_vec is not None:
+    # 2. Vector search (skipped when embeddings or sqlite-vec are unavailable)
+    if query_vec is not None and sqlite_vec is not None:
         nvec = top_k * 2
         try:
             if namespace:
@@ -564,8 +617,9 @@ def memory_forget(
     db = _get_conn()
     try:
         if col == "id":
-            # Vec table cascade
-            db.execute("DELETE FROM vec_memories WHERE id = ?", [id])
+            # Vec table cascade (absent when sqlite-vec is unavailable)
+            if sqlite_vec is not None:
+                db.execute("DELETE FROM vec_memories WHERE id = ?", [id])
             cur = db.execute("DELETE FROM memories WHERE id = ?", [id])
         else:
             cur = db.execute(sql, params)

package/src/pantheon/background-job-board.ts

@@ -2,7 +2,7 @@
  * Background Job Board — state machine for tracking background agent jobs.
  *
  * Tracks jobs through a state machine (running → completed/error/cancelled → reconciled),
- * supports alias generation per session, persistence via adapter, auto-wake signal files,
+ * supports alias generation per session, persistence via adapter, wake signal files,
  * and terminal state listeners (persist-before-notify write-ahead-log pattern).
  *
  * Pure TypeScript — zero external dependencies beyond Node.js builtins (fs/path).
@@ -123,7 +123,7 @@ export interface BoardOptions {
   /** Maximum completed jobs kept per agent for reuse (default: 3). */
   maxReusablePerAgent?: number
   /**
-   * Directory for auto-wake signal files.
+   * Directory for wake signal files.
    * When set, a `.signal.json` file is atomically written on every terminal transition.
    * When null or undefined, no signal files are written.
    */
@@ -175,7 +175,7 @@ function getAgentPrefix(agent: string): string {
  * - Strict state machine with validated transitions
  * - Per-session, per-agent alias generation (apo-1, her-2, …)
  * - Optional persistence via injected PersistenceAdapter
- * - Optional auto-wake signal files on terminal transitions
+ * - Optional wake signal files on terminal transitions
  * - Terminal state listeners with persist-before-notify guarantee
  * - expire/prune helpers for session lifecycle management
  */
@@ -652,7 +652,7 @@ export class BackgroundJobBoard {
   }
 
   /**
-   * Delete the auto-wake signal file for an alias (`<alias>.signal.json`).
+   * Delete the wake signal file for an alias (`<alias>.signal.json`).
    * No-op (false) when no signalDir is configured or the file is absent.
    * The delegate manager calls this on the spot right after auto-reconcile.
    */

package/src/pantheon/{delegation-compaction.ts → compaction-context.ts}

@@ -1,8 +1,8 @@
 /**
- * Compaction Context (Phase 4 + release-134 Phase 2) — build the context
- * blocks injected by the `experimental.session.compacting` hook so a
- * compacted session does not lose its working state: in-flight background
- * delegations, active goals and pending todos.
+ * Compaction Context (Phase 4 + release-134 Phase 2, slimmed in Fase 2 of the
+ * delegate removal) — build the context blocks injected by the
+ * `experimental.session.compacting` hook so a compacted session does not lose
+ * its working state: active goals and pending todos.
  *
  * Returns a Promise<string[]> (empty when there is nothing to preserve).
  * Section order is stable:
@@ -17,23 +17,10 @@
  *     `opts.todos`: "  [id] description — status". Omitted when the todo
  *     source is absent, disabled, unscoped (no sessionID), fails, or has
  *     no pending todos.
- *   - "Background Delegations (running):" — every running job launched by
- *     the session (active work is never capped): alias, agent, started ISO,
- *     truncated prompt.
- *   - "Background Delegations (finished, unread):" — terminal-unreconciled
- *     jobs, newest first, capped at `maxItems` (default 10, matching
- *     routing.yml background_delegation.max_compaction_items).
  *
- * Reconciled jobs are excluded — they have been read already.
- *
- * The block style mirrors board.formatForPrompt ("  [alias] description —
- * STATUS") so compaction context reads consistently with the system prompt,
- * but with the extra id/agent/started detail the compaction hook needs.
- *
- * @module delegation-compaction
+ * @module compaction-context
  */
 
-import type { BackgroundJobBoard, BackgroundJobRecord } from './background-job-board.ts'
 import type { Goal } from './goal-store.ts'
 import { createPantheonLogger } from './logger.ts'
 import type { TodoLike } from './todo-enforcer.ts'
@@ -72,56 +59,14 @@ export interface CompactionTodoSource {
 }
 
 export interface CompactionContextOptions {
-  /** Scope to jobs launched by this session; omit for all jobs. */
+  /** Scope to this session; omit to skip the per-session sections. */
   sessionID?: string
-  /** Cap on unread terminal jobs kept (default 10). */
-  maxItems?: number
   /** Goal source for <mission_context> (optional — omitted when absent). */
   goals?: CompactionGoalSource
   /** Todo source for <todo_context> (optional — omitted when absent). */
   todos?: CompactionTodoSource
 }
 
-/** Prompt truncation length for a context line. */
-const PROMPT_MAX_LEN = 100
-
-// ─── Formatting ────────────────────────────────────────────────────────
-
-/** Collapse whitespace and truncate a prompt to PROMPT_MAX_LEN chars. */
-function truncatePrompt(text: string): string {
-  const flat = text.trim().replace(/\s+/g, ' ')
-  if (flat.length <= PROMPT_MAX_LEN) return flat
-  return `${flat.slice(0, PROMPT_MAX_LEN)}…`
-}
-
-/** One-line terminal status label, mirroring formatForPrompt. */
-function terminalLabel(state: BackgroundJobRecord['state']): string {
-  switch (state) {
-    case 'completed':
-      return 'OK'
-    case 'error':
-      return 'ERR'
-    case 'cancelled':
-      return 'CAN'
-    default:
-      return state.toUpperCase()
-  }
-}
-
-function runningLine(job: BackgroundJobRecord): string {
-  return (
-    `  [${job.alias}] ${job.agent} — ${truncatePrompt(job.description)} — ` +
-    `RUNNING since ${new Date(job.launchedAt).toISOString()}`
-  )
-}
-
-function unreadTerminalLine(job: BackgroundJobRecord): string {
-  return (
-    `  [${job.alias}] ${job.agent} — ${truncatePrompt(job.description)} — ` +
-    `${terminalLabel(job.state)} ${new Date(job.updatedAt).toISOString()} [unread]`
-  )
-}
-
 // ─── Sections ──────────────────────────────────────────────────────────
 
 /**
@@ -180,58 +125,23 @@ async function todoContextBlock(opts: CompactionContextOptions): Promise<string
   return `<todo_context>\n${lines.join('\n')}`
 }
 
-/**
- * `<delegation_context>` — the existing delegation blocks, byte-for-byte
- * unchanged from Phase 4 (running, then unread terminal ≤ maxItems).
- */
-function delegationBlocks(board: BackgroundJobBoard, opts: CompactionContextOptions): string[] {
-  const maxItems = opts.maxItems ?? 10
-  const jobs = opts.sessionID !== undefined ? board.list(opts.sessionID) : board.list()
-
-  const running = jobs
-    .filter((j) => j.state === 'running')
-    .sort((a, b) => a.launchedAt - b.launchedAt)
-  const unreadTerminal = jobs
-    .filter((j) => j.terminalUnreconciled)
-    .sort((a, b) => b.updatedAt - a.updatedAt)
-    .slice(0, maxItems)
-
-  const blocks: string[] = []
-  if (running.length > 0) {
-    blocks.push(`Background Delegations (running):\n${running.map(runningLine).join('\n')}`)
-  }
-  if (unreadTerminal.length > 0) {
-    blocks.push(
-      `Background Delegations (finished, unread):\n${unreadTerminal
-        .map(unreadTerminalLine)
-        .join('\n')}`,
-    )
-  }
-  return blocks
-}
-
 // ─── Build ─────────────────────────────────────────────────────────────
 
 /**
  * Build the compaction context blocks for a session: preservation
- * directive, active goals, pending todos, and in-flight background
- * delegations. Returns an empty array when there is nothing to preserve
- * (caller skips the hook injection entirely).
+ * directive, active goals, and pending todos. Returns an empty array when
+ * there is nothing to preserve (caller skips the hook injection entirely).
  */
 export async function buildCompactionContext(
-  board: BackgroundJobBoard,
   opts: CompactionContextOptions = {},
 ): Promise<string[]> {
   const [mission, todo] = await Promise.all([missionContextBlock(opts), todoContextBlock(opts)])
-  const delegation = delegationBlocks(board, opts)
 
   // Totally-empty state → nothing to preserve; a lone directive is noise.
-  if (mission === undefined && todo === undefined && delegation.length === 0) return []
+  if (mission === undefined && todo === undefined) return []
 
-  const blocks: string[] = []
-  blocks.push(`<pantheon-context directive>\n${PANTHEON_COMPACTION_DIRECTIVE}`)
+  const blocks: string[] = [`<pantheon-context directive>\n${PANTHEON_COMPACTION_DIRECTIVE}`]
   if (mission !== undefined) blocks.push(mission)
   if (todo !== undefined) blocks.push(todo)
-  blocks.push(...delegation)
   return blocks
 }

package/src/pantheon/cost-command.ts

@@ -21,9 +21,8 @@ import { homedir } from 'node:os'
 import { delimiter, dirname, join } from 'node:path'
 import { fileURLToPath } from 'node:url'
 import { z } from 'zod'
-
-import type { ToolContextLike } from './delegation.ts'
 import type { NativeTaskStatus } from './native-task-status.ts'
+import type { ToolContextLike } from './tool-context.ts'
 
 const __filename = fileURLToPath(import.meta.url)
 const __dirname = dirname(__filename)