feat(memory): add grounded observation synthesis layer

agent/context_snapshots.py

@@ -1,69 +0,0 @@
-"""First-class context snapshot artifacts for live runtime memory evaluation."""
-
-from __future__ import annotations
-
-import json
-import re
-from pathlib import Path
-from typing import Any
-
-from hermes_constants import get_hermes_home
-
-
-_SAFE_SEGMENT_RE = re.compile(r"[^A-Za-z0-9_.-]+")
-
-
-class ContextSnapshotRecorder:
-    """Write per-call prompt-composition artifacts for a Hermes session."""
-
-    def __init__(self, session_id: str, *, enabled: bool = False, base_dir: str | Path | None = None):
-        self.session_id = session_id or "session"
-        self.enabled = bool(enabled)
-        self.base_dir = Path(base_dir) if base_dir else get_hermes_home() / "reports" / "context_snapshots"
-
-    @property
-    def session_dir(self) -> Path:
-        safe_session = _SAFE_SEGMENT_RE.sub("_", self.session_id).strip("._") or "session"
-        return self.base_dir / safe_session
-
-    def record_call(
-        self,
-        api_call_count: int,
-        *,
-        system_prompt: str,
-        memory_provider_system_prompt: str = "",
-        memory_prefetch_raw: str = "",
-        memory_context_block: str = "",
-        api_user_message: str = "",
-        api_messages: list[dict[str, Any]] | None = None,
-        metadata: dict[str, Any] | None = None,
-    ) -> Path | None:
-        if not self.enabled:
-            return None
-
-        call_dir = self.session_dir / f"call_{api_call_count:03d}"
-        call_dir.mkdir(parents=True, exist_ok=True)
-
-        self._write_text(call_dir / "system_prompt.txt", system_prompt or "")
-        self._write_text(call_dir / "memory_provider_system_prompt.txt", memory_provider_system_prompt or "")
-        self._write_text(call_dir / "memory_prefetch_raw.txt", memory_prefetch_raw or "")
-        self._write_text(call_dir / "memory_context_block.txt", memory_context_block or "")
-        self._write_text(call_dir / "api_user_message.txt", api_user_message or "")
-        self._write_json(call_dir / "api_messages.json", api_messages or [])
-        self._write_json(
-            call_dir / "metadata.json",
-            {
-                "session_id": self.session_id,
-                "api_call_count": api_call_count,
-                **(metadata or {}),
-            },
-        )
-        return call_dir
-
-    @staticmethod
-    def _write_text(path: Path, content: str) -> None:
-        path.write_text(content, encoding="utf-8")
-
-    @staticmethod
-    def _write_json(path: Path, payload: Any) -> None:
-        path.write_text(json.dumps(payload, indent=2, ensure_ascii=False), encoding="utf-8")

docs/hindsight-local-eval.md

@@ -1,132 +0,0 @@
-# Hindsight local eval homes for live Hermes runtime testing
-
-Issue: #1010
-Parent: #985
-
-This document defines a reproducible, profile-scoped evaluation layout for baseline / MemPalace / Hindsight comparisons without requiring Hindsight Cloud.
-
-## Eval home layout
-
-Use three separate `HERMES_HOME` directories so each run has isolated config, memory, sessions, and artifacts.
-
-```text
-~/.hermes/profiles/atlas-baseline/
-  config.yaml
-  .env
-  MEMORY.md
-  USER.md
-  reports/context_snapshots/
-
-~/.hermes/profiles/atlas-mempalace/
-  config.yaml
-  .env
-  MEMORY.md
-  USER.md
-  reports/context_snapshots/
-  plugins/           # if a local MemPalace plugin is installed for this eval lane
-
-~/.hermes/profiles/atlas-hindsight/
-  config.yaml
-  .env
-  MEMORY.md
-  USER.md
-  hindsight/config.json
-  reports/context_snapshots/
-```
-
-## Hindsight local config
-
-The Hindsight provider already loads config from `$HERMES_HOME/hindsight/config.json` first. For the local eval lane, prefer `local_embedded` so Hermes can bring up a local Hindsight daemon without cloud signup.
-
-Example `~/.hermes/profiles/atlas-hindsight/hindsight/config.json`:
-
-```json
-{
-  "mode": "local_embedded",
-  "memory_mode": "context",
-  "recall_prefetch_method": "recall",
-  "llm_provider": "ollama",
-  "llm_model": "gemma3:12b",
-  "api_url": "http://localhost:8888"
-}
-```
-
-Notes:
-- `local_embedded` avoids any Hindsight Cloud dependency.
-- If `profile` is omitted, Hermes now derives a stable local Hindsight profile name from the active profile identity / `HERMES_HOME` instead of collapsing all local runs into the shared legacy `hermes` profile.
-- `local_external` remains valid if you already run a local Hindsight server yourself.
-
-## Runtime switching procedure
-
-Switch by exporting `HERMES_HOME` before launching Hermes.
-
-### 1. Baseline
-
-```bash
-export HERMES_HOME="$HOME/.hermes/profiles/atlas-baseline"
-unset HERMES_CONTEXT_SNAPSHOTS
-hermes chat
-```
-
-### 2. MemPalace lane
-
-```bash
-export HERMES_HOME="$HOME/.hermes/profiles/atlas-mempalace"
-export HERMES_CONTEXT_SNAPSHOTS=1
-hermes chat
-```
-
-### 3. Hindsight lane
-
-```bash
-export HERMES_HOME="$HOME/.hermes/profiles/atlas-hindsight"
-export HERMES_CONTEXT_SNAPSHOTS=1
-hermes chat
-```
-
-## Raw artifact capture
-
-When `HERMES_CONTEXT_SNAPSHOTS=1` is enabled, Hermes writes first-class prompt-composition artifacts under the active home by default.
-
-Artifact tree:
-
-```text
-$HERMES_HOME/reports/context_snapshots/<session-id>/call_001/
-  system_prompt.txt
-  memory_provider_system_prompt.txt
-  memory_prefetch_raw.txt
-  memory_context_block.txt
-  api_user_message.txt
-  api_messages.json
-  metadata.json
-```
-
-Minimum files a benchmark should inspect:
-- `system_prompt.txt`
-- `memory_prefetch_raw.txt`
-- `memory_context_block.txt`
-- `api_user_message.txt`
-- `api_messages.json`
-
-These prove:
-- what the system prompt was
-- what the provider prefetched
-- what entered `<memory-context>`
-- what the final API user message looked like
-- what full payload reached the model
-
-## Follow-on benchmark workflow
-
-A benchmark issue can now consume this path without redoing integration work:
-1. pick one eval home (`atlas-baseline`, `atlas-mempalace`, `atlas-hindsight`)
-2. export the corresponding `HERMES_HOME`
-3. run Hermes on the same prompt set
-4. compare the snapshot artifacts in `reports/context_snapshots/`
-5. score recall quality and answer quality separately
-
-## Why this is sovereign
-
-- no hosted Hindsight Cloud dependency is required
-- the Hindsight config is profile-scoped under `hindsight/config.json`
-- the runtime artifacts stay under the active `HERMES_HOME`
-- switching between baseline / MemPalace / Hindsight is just a `HERMES_HOME` swap

plugins/memory/hindsight/__init__.py

@@ -178,25 +178,6 @@ def _load_config() -> dict:
     }
 
 
-def _derive_local_profile_name(agent_identity: str = "", hermes_home: str = "") -> str:
-    """Return a stable profile name for local embedded Hindsight storage.
-
-    Prefer the active Hermes profile identity when available, otherwise fall back
-    to the basename of the active HERMES_HOME path. This prevents all local
-    Hindsight eval homes from sharing the legacy default profile name "hermes".
-    """
-    from pathlib import Path
-    import re
-
-    raw = (agent_identity or "").strip()
-    if not raw and hermes_home:
-        raw = Path(hermes_home).name.strip()
-    if not raw:
-        raw = "hermes"
-    safe = re.sub(r"[^A-Za-z0-9_.-]+", "-", raw).strip(".-_")
-    return safe or "hermes"
-
-
 # ---------------------------------------------------------------------------
 # MemoryProvider implementation
 # ---------------------------------------------------------------------------
@@ -487,8 +468,6 @@ class HindsightMemoryProvider(MemoryProvider):
 
     def initialize(self, session_id: str, **kwargs) -> None:
         self._session_id = session_id
-        hermes_home = str(kwargs.get("hermes_home") or "")
-        agent_identity = str(kwargs.get("agent_identity") or "")
 
         # Check client version and auto-upgrade if needed
         try:
@@ -521,11 +500,6 @@ class HindsightMemoryProvider(MemoryProvider):
         # "local" is a legacy alias for "local_embedded"
         if self._mode == "local":
             self._mode = "local_embedded"
-        if self._mode == "local_embedded" and not self._config.get("profile"):
-            self._config["profile"] = _derive_local_profile_name(
-                agent_identity=agent_identity,
-                hermes_home=hermes_home,
-            )
         self._api_key = self._config.get("apiKey") or self._config.get("api_key") or os.environ.get("HINDSIGHT_API_KEY", "")
         default_url = _DEFAULT_LOCAL_URL if self._mode in ("local_embedded", "local_external") else _DEFAULT_API_URL
         self._api_url = self._config.get("api_url") or os.environ.get("HINDSIGHT_API_URL", default_url)

plugins/memory/holographic/__init__.py

@@ -26,6 +26,7 @@ from agent.memory_provider import MemoryProvider
 from tools.registry import tool_error
 from .store import MemoryStore
 from .retrieval import FactRetriever
+from .observations import ObservationSynthesizer
 
 logger = logging.getLogger(__name__)
 
@@ -37,28 +38,29 @@ logger = logging.getLogger(__name__)
 FACT_STORE_SCHEMA = {
     "name": "fact_store",
     "description": (
-        "Deep structured memory with algebraic reasoning. "
+        "Deep structured memory with algebraic reasoning and grounded observation synthesis. "
         "Use alongside the memory tool — memory for always-on context, "
-        "fact_store for deep recall and compositional queries.\n\n"
+        "fact_store for deep recall, compositional queries, and higher-order observations.\n\n"
         "ACTIONS (simple → powerful):\n"
         "• add — Store a fact the user would expect you to remember.\n"
         "• search — Keyword lookup ('editor config', 'deploy process').\n"
         "• probe — Entity recall: ALL facts about a person/thing.\n"
         "• related — What connects to an entity? Structural adjacency.\n"
         "• reason — Compositional: facts connected to MULTIPLE entities simultaneously.\n"
+        "• observe — Synthesized higher-order observations backed by supporting facts.\n"
         "• contradict — Memory hygiene: find facts making conflicting claims.\n"
         "• update/remove/list — CRUD operations.\n\n"
-        "IMPORTANT: Before answering questions about the user, ALWAYS probe or reason first."
+        "IMPORTANT: Before answering questions about the user, ALWAYS probe/reason/observe first."
     ),
     "parameters": {
         "type": "object",
         "properties": {
             "action": {
                 "type": "string",
-                "enum": ["add", "search", "probe", "related", "reason", "contradict", "update", "remove", "list"],
+                "enum": ["add", "search", "probe", "related", "reason", "observe", "contradict", "update", "remove", "list"],
             },
             "content": {"type": "string", "description": "Fact content (required for 'add')."},
-            "query": {"type": "string", "description": "Search query (required for 'search')."},
+            "query": {"type": "string", "description": "Search query (required for 'search'/'observe')."},
             "entity": {"type": "string", "description": "Entity name for 'probe'/'related'."},
             "entities": {"type": "array", "items": {"type": "string"}, "description": "Entity names for 'reason'."},
             "fact_id": {"type": "integer", "description": "Fact ID for 'update'/'remove'."},
@@ -66,6 +68,12 @@ FACT_STORE_SCHEMA = {
             "tags": {"type": "string", "description": "Comma-separated tags."},
             "trust_delta": {"type": "number", "description": "Trust adjustment for 'update'."},
             "min_trust": {"type": "number", "description": "Minimum trust filter (default: 0.3)."},
+            "min_confidence": {"type": "number", "description": "Minimum observation confidence (default: 0.6)."},
+            "observation_type": {
+                "type": "string",
+                "enum": ["recurring_preference", "stable_direction", "behavioral_pattern"],
+                "description": "Optional observation type filter for 'observe'.",
+            },
             "limit": {"type": "integer", "description": "Max results (default: 10)."},
         },
         "required": ["action"],
@@ -118,7 +126,9 @@ class HolographicMemoryProvider(MemoryProvider):
         self._config = config or _load_plugin_config()
         self._store = None
         self._retriever = None
+        self._observation_synth = None
         self._min_trust = float(self._config.get("min_trust_threshold", 0.3))
+        self._observation_min_confidence = float(self._config.get("observation_min_confidence", 0.6))
 
     @property
     def name(self) -> str:
@@ -177,6 +187,7 @@ class HolographicMemoryProvider(MemoryProvider):
             hrr_weight=hrr_weight,
             hrr_dim=hrr_dim,
         )
+        self._observation_synth = ObservationSynthesizer(self._store)
         self._session_id = session_id
 
     def system_prompt_block(self) -> str:
@@ -193,30 +204,76 @@ class HolographicMemoryProvider(MemoryProvider):
                 "# Holographic Memory\n"
                 "Active. Empty fact store — proactively add facts the user would expect you to remember.\n"
                 "Use fact_store(action='add') to store durable structured facts about people, projects, preferences, decisions.\n"
+                "Use fact_store(action='observe') to synthesize higher-order observations with evidence.\n"
                 "Use fact_feedback to rate facts after using them (trains trust scores)."
             )
         return (
             f"# Holographic Memory\n"
             f"Active. {total} facts stored with entity resolution and trust scoring.\n"
-            f"Use fact_store to search, probe entities, reason across entities, or add facts.\n"
+            f"Use fact_store to search, probe entities, reason across entities, or synthesize observations.\n"
             f"Use fact_feedback to rate facts after using them (trains trust scores)."
         )
 
     def prefetch(self, query: str, *, session_id: str = "") -> str:
-        if not self._retriever or not query:
+        if not query:
             return ""
+
+        parts = []
+        raw_results = []
         try:
-            results = self._retriever.search(query, min_trust=self._min_trust, limit=5)
-            if not results:
-                return ""
+            if self._retriever:
+                raw_results = self._retriever.search(query, min_trust=self._min_trust, limit=5)
+        except Exception as e:
+            logger.debug("Holographic prefetch fact search failed: %s", e)
+            raw_results = []
+
+        observations = []
+        try:
+            if self._observation_synth:
+                observations = self._observation_synth.observe(
+                    query,
+                    min_confidence=self._observation_min_confidence,
+                    limit=3,
+                    refresh=True,
+                )
+        except Exception as e:
+            logger.debug("Holographic prefetch observation search failed: %s", e)
+            observations = []
+
+        if not raw_results and observations:
+            seen_fact_ids = set()
+            evidence_backfill = []
+            for observation in observations:
+                for evidence in observation.get("evidence", []):
+                    fact_id = evidence.get("fact_id")
+                    if fact_id in seen_fact_ids:
+                        continue
+                    seen_fact_ids.add(fact_id)
+                    evidence_backfill.append(evidence)
+            raw_results = evidence_backfill[:5]
+
+        if raw_results:
             lines = []
-            for r in results:
+            for r in raw_results:
                 trust = r.get("trust_score", r.get("trust", 0))
                 lines.append(f"- [{trust:.1f}] {r.get('content', '')}")
-            return "## Holographic Memory\n" + "\n".join(lines)
-        except Exception as e:
-            logger.debug("Holographic prefetch failed: %s", e)
-            return ""
+            parts.append("## Holographic Memory\n" + "\n".join(lines))
+
+        if observations:
+            lines = []
+            for observation in observations:
+                evidence_ids = ", ".join(
+                    f"#{item['fact_id']}" for item in observation.get("evidence", [])[:3]
+                ) or "none"
+                lines.append(
+                    f"- [{observation.get('confidence', 0.0):.2f}] "
+                    f"{observation.get('observation_type', 'observation')}: "
+                    f"{observation.get('summary', '')} "
+                    f"(evidence: {evidence_ids})"
+                )
+            parts.append("## Holographic Observations\n" + "\n".join(lines))
+
+        return "\n\n".join(parts)
 
     def sync_turn(self, user_content: str, assistant_content: str, *, session_id: str = "") -> None:
         # Holographic memory stores explicit facts via tools, not auto-sync.
@@ -252,6 +309,7 @@ class HolographicMemoryProvider(MemoryProvider):
     def shutdown(self) -> None:
         self._store = None
         self._retriever = None
+        self._observation_synth = None
 
     # -- Tool handlers -------------------------------------------------------
 
@@ -305,6 +363,19 @@ class HolographicMemoryProvider(MemoryProvider):
                 )
                 return json.dumps({"results": results, "count": len(results)})
 
+            elif action == "observe":
+                synthesizer = self._observation_synth
+                if not synthesizer:
+                    return tool_error("Observation synthesizer is not initialized")
+                observations = synthesizer.observe(
+                    args.get("query", ""),
+                    observation_type=args.get("observation_type"),
+                    min_confidence=float(args.get("min_confidence", self._observation_min_confidence)),
+                    limit=int(args.get("limit", 10)),
+                    refresh=True,
+                )
+                return json.dumps({"observations": observations, "count": len(observations)})
+
             elif action == "contradict":
                 results = retriever.contradict(
                     category=args.get("category"),

plugins/memory/holographic/observations.py

@@ -0,0 +1,249 @@
+"""Higher-order observation synthesis for holographic memory.
+
+Builds grounded observations from accumulated facts and keeps them in a
+separate retrieval layer with explicit evidence links back to supporting facts.
+"""
+
+from __future__ import annotations
+
+import re
+from typing import Any
+
+from .store import MemoryStore
+
+_TOKEN_RE = re.compile(r"[a-z0-9_]+")
+_HIGHER_ORDER_CUES = {
+    "prefer",
+    "preference",
+    "preferences",
+    "style",
+    "pattern",
+    "patterns",
+    "behavior",
+    "behaviour",
+    "habit",
+    "habits",
+    "workflow",
+    "direction",
+    "trajectory",
+    "strategy",
+    "tend",
+    "usually",
+}
+
+_OBSERVATION_PATTERNS = [
+    {
+        "observation_type": "recurring_preference",
+        "subject": "communication_style",
+        "categories": {"user_pref", "general"},
+        "labels": {
+            "concise": ["concise", "terse", "brief", "short", "no fluff"],
+            "result_first": ["result-only", "result only", "outcome only", "quick", "quickly"],
+            "silent_ops": ["silent", "no status", "no repetitive status", "no questions"],
+        },
+        "summary_prefix": "Recurring preference",
+    },
+    {
+        "observation_type": "stable_direction",
+        "subject": "project_direction",
+        "categories": {"project", "general", "tool"},
+        "labels": {
+            "local_first": ["local-first", "local first", "local-only", "local only", "ollama", "own hardware"],
+            "gitea_first": ["gitea-first", "gitea first", "forge", "pull request", "pr flow", "issue flow"],
+            "ansible": ["ansible", "playbook", "role", "deploy via ansible"],
+        },
+        "summary_prefix": "Stable direction",
+    },
+    {
+        "observation_type": "behavioral_pattern",
+        "subject": "operator_workflow",
+        "categories": {"general", "project", "tool", "user_pref"},
+        "labels": {
+            "commit_early": ["commit early", "commits early", "commit after", "wip commit"],
+            "pr_first": ["open pr", "push a pr", "pull request", "pr immediately", "create pr"],
+            "dedup_guard": ["no dupes", "no duplicates", "avoid duplicate", "existing pr"],
+        },
+        "summary_prefix": "Behavioral pattern",
+    },
+]
+
+_TYPE_QUERY_HINTS = {
+    "recurring_preference": {"prefer", "preference", "style", "communication", "likes", "wants"},
+    "stable_direction": {"direction", "trajectory", "strategy", "project", "roadmap", "moving"},
+    "behavioral_pattern": {"pattern", "behavior", "workflow", "habit", "operator", "agent", "usually"},
+}
+
+
+class ObservationSynthesizer:
+    """Synthesizes grounded observations from facts and retrieves them by query."""
+
+    def __init__(self, store: MemoryStore):
+        self.store = store
+
+    def synthesize(
+        self,
+        *,
+        persist: bool = True,
+        min_confidence: float = 0.6,
+        limit: int = 10,
+    ) -> list[dict[str, Any]]:
+        facts = self.store.list_facts(min_trust=0.0, limit=1000)
+        observations: list[dict[str, Any]] = []
+
+        for pattern in _OBSERVATION_PATTERNS:
+            candidate = self._build_candidate(pattern, facts, min_confidence=min_confidence)
+            if not candidate:
+                continue
+
+            if persist:
+                candidate["observation_id"] = self.store.upsert_observation(
+                    candidate["observation_type"],
+                    candidate["subject"],
+                    candidate["summary"],
+                    candidate["confidence"],
+                    candidate["evidence_fact_ids"],
+                    metadata=candidate["metadata"],
+                )
+
+            candidate["evidence"] = self._expand_evidence(candidate["evidence_fact_ids"])
+            candidate["evidence_count"] = len(candidate["evidence"])
+            candidate.pop("evidence_fact_ids", None)
+            observations.append(candidate)
+
+        observations.sort(
+            key=lambda item: (item["confidence"], item.get("evidence_count", 0)),
+            reverse=True,
+        )
+        return observations[:limit]
+
+    def observe(
+        self,
+        query: str = "",
+        *,
+        observation_type: str | None = None,
+        min_confidence: float = 0.6,
+        limit: int = 10,
+        refresh: bool = True,
+    ) -> list[dict[str, Any]]:
+        if refresh:
+            self.synthesize(persist=True, min_confidence=min_confidence, limit=limit)
+
+        observations = self.store.list_observations(
+            observation_type=observation_type,
+            min_confidence=min_confidence,
+            limit=max(limit * 4, 20),
+        )
+        if not observations:
+            return []
+
+        if not query:
+            return observations[:limit]
+
+        query_tokens = self._tokenize(query)
+        is_higher_order = bool(query_tokens & _HIGHER_ORDER_CUES)
+        ranked: list[dict[str, Any]] = []
+
+        for item in observations:
+            searchable = " ".join(
+                [
+                    item.get("summary", ""),
+                    item.get("subject", ""),
+                    item.get("observation_type", ""),
+                    " ".join(item.get("metadata", {}).get("labels", [])),
+                ]
+            )
+            overlap = self._overlap_score(query_tokens, self._tokenize(searchable))
+            type_bonus = self._type_bonus(query_tokens, item.get("observation_type", ""))
+            if overlap <= 0 and type_bonus <= 0 and not is_higher_order:
+                continue
+            ranked_item = dict(item)
+            ranked_item["score"] = round(item.get("confidence", 0.0) + overlap + type_bonus, 3)
+            ranked.append(ranked_item)
+
+        if not ranked and is_higher_order:
+            ranked = [
+                {**item, "score": round(float(item.get("confidence", 0.0)), 3)}
+                for item in observations
+            ]
+
+        ranked.sort(
+            key=lambda item: (item.get("score", 0.0), item.get("confidence", 0.0), item.get("evidence_count", 0)),
+            reverse=True,
+        )
+        return ranked[:limit]
+
+    def _build_candidate(
+        self,
+        pattern: dict[str, Any],
+        facts: list[dict[str, Any]],
+        *,
+        min_confidence: float,
+    ) -> dict[str, Any] | None:
+        matched_fact_ids: set[int] = set()
+        matched_labels: dict[str, set[int]] = {label: set() for label in pattern["labels"]}
+
+        for fact in facts:
+            if fact.get("category") not in pattern["categories"]:
+                continue
+            haystack = f"{fact.get('content', '')} {fact.get('tags', '')}".lower()
+            local_match = False
+            for label, keywords in pattern["labels"].items():
+                if any(keyword in haystack for keyword in keywords):
+                    matched_labels[label].add(int(fact["fact_id"]))
+                    local_match = True
+            if local_match:
+                matched_fact_ids.add(int(fact["fact_id"]))
+
+        if len(matched_fact_ids) < 2:
+            return None
+
+        active_labels = sorted(label for label, ids in matched_labels.items() if ids)
+        confidence = min(0.95, 0.35 + 0.12 * len(matched_fact_ids) + 0.08 * len(active_labels))
+        confidence = round(confidence, 3)
+        if confidence < min_confidence:
+            return None
+
+        label_summary = ", ".join(label.replace("_", "-") for label in active_labels)
+        subject_text = pattern["subject"].replace("_", " ")
+        summary = (
+            f"{pattern['summary_prefix']}: {subject_text} trends toward {label_summary} "
+            f"based on {len(matched_fact_ids)} supporting facts."
+        )
+        return {
+            "observation_type": pattern["observation_type"],
+            "subject": pattern["subject"],
+            "summary": summary,
+            "confidence": confidence,
+            "metadata": {
+                "labels": active_labels,
+                "evidence_count": len(matched_fact_ids),
+            },
+            "evidence_fact_ids": sorted(matched_fact_ids),
+        }
+
+    def _expand_evidence(self, fact_ids: list[int]) -> list[dict[str, Any]]:
+        facts_by_id = {
+            fact["fact_id"]: fact
+            for fact in self.store.list_facts(min_trust=0.0, limit=1000)
+        }
+        return [facts_by_id[fact_id] for fact_id in fact_ids if fact_id in facts_by_id]
+
+    @staticmethod
+    def _tokenize(text: str) -> set[str]:
+        return set(_TOKEN_RE.findall(text.lower()))
+
+    @staticmethod
+    def _overlap_score(query_tokens: set[str], text_tokens: set[str]) -> float:
+        if not query_tokens or not text_tokens:
+            return 0.0
+        overlap = query_tokens & text_tokens
+        if not overlap:
+            return 0.0
+        return round(len(overlap) / max(len(query_tokens), 1), 3)
+
+    @staticmethod
+    def _type_bonus(query_tokens: set[str], observation_type: str) -> float:
+        hints = _TYPE_QUERY_HINTS.get(observation_type, set())
+        if not hints:
+            return 0.0
+        return 0.25 if query_tokens & hints else 0.0

plugins/memory/holographic/store.py

@@ -3,6 +3,7 @@ SQLite-backed fact store with entity resolution and trust scoring.
 Single-user Hermes memory store plugin.
 """
 
+import json
 import re
 import sqlite3
 import threading
@@ -73,6 +74,28 @@ CREATE TABLE IF NOT EXISTS memory_banks (
     fact_count INTEGER DEFAULT 0,
     updated_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP
 );
+
+CREATE TABLE IF NOT EXISTS observations (
+    observation_id   INTEGER PRIMARY KEY AUTOINCREMENT,
+    observation_type TEXT NOT NULL,
+    subject          TEXT NOT NULL,
+    summary          TEXT NOT NULL,
+    confidence       REAL DEFAULT 0.0,
+    metadata_json    TEXT DEFAULT '{}',
+    created_at       TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
+    updated_at       TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
+    UNIQUE(observation_type, subject)
+);
+
+CREATE TABLE IF NOT EXISTS observation_evidence (
+    observation_id INTEGER REFERENCES observations(observation_id) ON DELETE CASCADE,
+    fact_id        INTEGER REFERENCES facts(fact_id) ON DELETE CASCADE,
+    evidence_weight REAL DEFAULT 1.0,
+    PRIMARY KEY (observation_id, fact_id)
+);
+
+CREATE INDEX IF NOT EXISTS idx_observations_type ON observations(observation_type);
+CREATE INDEX IF NOT EXISTS idx_observations_confidence ON observations(confidence DESC);
 """
 
 # Trust adjustment constants
@@ -128,6 +151,7 @@ class MemoryStore:
     def _init_db(self) -> None:
         """Create tables, indexes, and triggers if they do not exist. Enable WAL mode."""
         self._conn.execute("PRAGMA journal_mode=WAL")
+        self._conn.execute("PRAGMA foreign_keys=ON")
         self._conn.executescript(_SCHEMA)
         # Migrate: add hrr_vector column if missing (safe for existing databases)
         columns = {row[1] for row in self._conn.execute("PRAGMA table_info(facts)").fetchall()}
@@ -346,6 +370,115 @@ class MemoryStore:
             rows = self._conn.execute(sql, params).fetchall()
             return [self._row_to_dict(r) for r in rows]
 
+    def upsert_observation(
+        self,
+        observation_type: str,
+        subject: str,
+        summary: str,
+        confidence: float,
+        evidence_fact_ids: list[int],
+        metadata: dict | None = None,
+    ) -> int:
+        """Create or update a synthesized observation and its evidence links."""
+        with self._lock:
+            metadata_json = json.dumps(metadata or {}, sort_keys=True)
+            self._conn.execute(
+                """
+                INSERT INTO observations (
+                    observation_type, subject, summary, confidence, metadata_json
+                )
+                VALUES (?, ?, ?, ?, ?)
+                ON CONFLICT(observation_type, subject) DO UPDATE SET
+                    summary = excluded.summary,
+                    confidence = excluded.confidence,
+                    metadata_json = excluded.metadata_json,
+                    updated_at = CURRENT_TIMESTAMP
+                """,
+                (observation_type, subject, summary, confidence, metadata_json),
+            )
+            row = self._conn.execute(
+                """
+                SELECT observation_id
+                FROM observations
+                WHERE observation_type = ? AND subject = ?
+                """,
+                (observation_type, subject),
+            ).fetchone()
+            observation_id = int(row["observation_id"])
+
+            self._conn.execute(
+                "DELETE FROM observation_evidence WHERE observation_id = ?",
+                (observation_id,),
+            )
+            unique_fact_ids = sorted({int(fid) for fid in evidence_fact_ids})
+            if unique_fact_ids:
+                self._conn.executemany(
+                    """
+                    INSERT OR IGNORE INTO observation_evidence (observation_id, fact_id)
+                    VALUES (?, ?)
+                    """,
+                    [(observation_id, fact_id) for fact_id in unique_fact_ids],
+                )
+            self._conn.commit()
+            return observation_id
+
+    def list_observations(
+        self,
+        observation_type: str | None = None,
+        min_confidence: float = 0.0,
+        limit: int = 50,
+    ) -> list[dict]:
+        """List synthesized observations with expanded supporting evidence."""
+        with self._lock:
+            params: list = [min_confidence]
+            observation_clause = ""
+            if observation_type is not None:
+                observation_clause = "AND observation_type = ?"
+                params.append(observation_type)
+            params.append(limit)
+            rows = self._conn.execute(
+                f"""
+                SELECT observation_id, observation_type, subject, summary, confidence,
+                       metadata_json, created_at, updated_at,
+                       (
+                           SELECT COUNT(*)
+                           FROM observation_evidence oe
+                           WHERE oe.observation_id = observations.observation_id
+                       ) AS evidence_count
+                FROM observations
+                WHERE confidence >= ?
+                  {observation_clause}
+                ORDER BY confidence DESC, updated_at DESC
+                LIMIT ?
+                """,
+                params,
+            ).fetchall()
+
+            results = []
+            for row in rows:
+                item = dict(row)
+                try:
+                    item["metadata"] = json.loads(item.pop("metadata_json") or "{}")
+                except json.JSONDecodeError:
+                    item["metadata"] = {}
+                item["evidence"] = self._get_observation_evidence(int(item["observation_id"]))
+                results.append(item)
+            return results
+
+    def _get_observation_evidence(self, observation_id: int) -> list[dict]:
+        rows = self._conn.execute(
+            """
+            SELECT f.fact_id, f.content, f.category, f.tags, f.trust_score,
+                   f.retrieval_count, f.helpful_count, f.created_at, f.updated_at
+            FROM observation_evidence oe
+            JOIN facts f ON f.fact_id = oe.fact_id
+            WHERE oe.observation_id = ?
+            ORDER BY f.trust_score DESC, f.updated_at DESC
+            """,
+            (observation_id,),
+        ).fetchall()
+        return [self._row_to_dict(row) for row in rows]
+
     def record_feedback(self, fact_id: int, helpful: bool) -> dict:
         """Record user feedback and adjust trust asymmetrically.
 

run_agent.py

@@ -604,8 +604,6 @@ class AIAgent:
         checkpoint_max_snapshots: int = 50,
         pass_session_id: bool = False,
         persist_session: bool = True,
-        context_snapshots_enabled: bool | None = None,
-        context_snapshots_dir: str | None = None,
     ):
         """
         Initialize the AI Agent.
@@ -1131,43 +1129,6 @@ class AIAgent:
         except Exception:
             _agent_cfg = {}
 
-        def _is_enabled(value):
-            if isinstance(value, bool):
-                return value
-            return str(value).strip().lower() in {"1", "true", "yes", "on"}
-
-        _debug_cfg = _agent_cfg.get("debug", {}) if isinstance(_agent_cfg, dict) else {}
-        if not isinstance(_debug_cfg, dict):
-            _debug_cfg = {}
-        _snapshot_cfg = _debug_cfg.get("context_snapshots", {})
-        if not isinstance(_snapshot_cfg, dict):
-            _snapshot_cfg = {}
-        _snapshots_env = os.getenv("HERMES_CONTEXT_SNAPSHOTS")
-        _snapshots_dir_env = os.getenv("HERMES_CONTEXT_SNAPSHOTS_DIR")
-        if context_snapshots_enabled is None:
-            if _snapshots_env is not None:
-                self._context_snapshots_enabled = _is_enabled(_snapshots_env)
-            else:
-                self._context_snapshots_enabled = _is_enabled(_snapshot_cfg.get("enabled", False))
-        else:
-            self._context_snapshots_enabled = bool(context_snapshots_enabled)
-        self._context_snapshots_dir = (
-            context_snapshots_dir
-            or _snapshots_dir_env
-            or _snapshot_cfg.get("dir")
-            or None
-        )
-        try:
-            from agent.context_snapshots import ContextSnapshotRecorder
-            self._context_snapshot_recorder = ContextSnapshotRecorder(
-                session_id=self.session_id,
-                enabled=self._context_snapshots_enabled,
-                base_dir=self._context_snapshots_dir,
-            )
-        except Exception as _snapshot_err:
-            logger.debug("Context snapshot recorder init failed: %s", _snapshot_err)
-            self._context_snapshot_recorder = None
-
         # Persistent memory (MEMORY.md + USER.md) -- loaded from disk
         self._memory_store = None
         self._memory_enabled = False
@@ -8183,17 +8144,12 @@ class AIAgent:
         # Use original_user_message (clean input) — user_message may contain
         # injected skill content that bloats / breaks provider queries.
         _ext_prefetch_cache = ""
-        _memory_provider_prompt_cache = ""
         if self._memory_manager:
             try:
                 _query = original_user_message if isinstance(original_user_message, str) else ""
                 _ext_prefetch_cache = self._memory_manager.prefetch_all(_query) or ""
             except Exception:
                 pass
-            try:
-                _memory_provider_prompt_cache = self._memory_manager.build_system_prompt() or ""
-            except Exception:
-                pass
 
         while (api_call_count < self.max_iterations and self.iteration_budget.remaining > 0) or self._budget_grace_call:
             # Reset per-turn checkpoint dedup so each iteration can take one snapshot
@@ -8261,8 +8217,6 @@ class AIAgent:
             # However, providers like Moonshot AI require a separate 'reasoning_content' field
             # on assistant messages with tool_calls. We handle both cases here.
             api_messages = []
-            _current_api_user_message = ""
-            _current_memory_context_block = ""
             for idx, msg in enumerate(messages):
                 api_msg = msg.copy()
 
@@ -8277,15 +8231,12 @@ class AIAgent:
                         _fenced = build_memory_context_block(_ext_prefetch_cache)
                         if _fenced:
                             _injections.append(_fenced)
-                            _current_memory_context_block = _fenced
                     if _plugin_user_context:
                         _injections.append(_plugin_user_context)
                     if _injections:
                         _base = api_msg.get("content", "")
                         if isinstance(_base, str):
                             api_msg["content"] = _base + "\n\n" + "\n\n".join(_injections)
-                    if isinstance(api_msg.get("content"), str):
-                        _current_api_user_message = api_msg["content"]
 
                 # For ALL assistant messages, pass reasoning back to the API
                 # This ensures multi-turn reasoning context is preserved
@@ -8320,13 +8271,7 @@ class AIAgent:
                 from agent.privacy_filter import PrivacyFilter
                 pf = PrivacyFilter()
                 # Sanitize messages before they reach the provider
-                _pf_result = pf.sanitize_messages(api_messages)
-                if isinstance(_pf_result, tuple):
-                    api_messages, _pf_report = _pf_result
-                    if getattr(pf, "last_report", None) is None:
-                        pf.last_report = _pf_report
-                else:
-                    api_messages = _pf_result
+                api_messages = pf.sanitize_messages(api_messages)
                 if pf.last_report and pf.last_report.had_redactions:
                     logger.info(f"Privacy Filter: Redacted sensitive data from turn payload. Details: {pf.last_report.summary()}")
             except Exception as e:
@@ -8397,27 +8342,6 @@ class AIAgent:
                     new_tcs.append(tc)
                 am["tool_calls"] = new_tcs
 
-            if self._context_snapshot_recorder:
-                try:
-                    self._context_snapshot_recorder.record_call(
-                        api_call_count,
-                        system_prompt=effective_system,
-                        memory_provider_system_prompt=_memory_provider_prompt_cache,
-                        memory_prefetch_raw=_ext_prefetch_cache,
-                        memory_context_block=_current_memory_context_block,
-                        api_user_message=_current_api_user_message,
-                        api_messages=api_messages,
-                        metadata={
-                            "model": self.model,
-                            "provider": self.provider,
-                            "platform": self.platform or "",
-                            "api_mode": self.api_mode,
-                            "memory_providers": [p.name for p in getattr(self._memory_manager, "providers", [])],
-                        },
-                    )
-                except Exception as _snapshot_err:
-                    logger.debug("Context snapshot capture failed: %s", _snapshot_err)
-
             # Calculate approximate request size for logging
             total_chars = sum(len(str(msg)) for msg in api_messages)
             approx_tokens = estimate_messages_tokens_rough(api_messages)

tests/agent/test_context_snapshots.py

@@ -1,43 +0,0 @@
-from pathlib import Path
-
-from agent.context_snapshots import ContextSnapshotRecorder
-
-
-def test_disabled_recorder_writes_nothing(tmp_path):
-    recorder = ContextSnapshotRecorder(session_id="session-1", enabled=False, base_dir=tmp_path)
-
-    out = recorder.record_call(
-        1,
-        system_prompt="system",
-        api_messages=[{"role": "user", "content": "hello"}],
-    )
-
-    assert out is None
-    assert not (tmp_path / "session-1").exists()
-
-
-def test_enabled_recorder_writes_expected_artifacts(tmp_path):
-    recorder = ContextSnapshotRecorder(session_id="session-1", enabled=True, base_dir=tmp_path)
-
-    out = recorder.record_call(
-        1,
-        system_prompt="system prompt",
-        memory_provider_system_prompt="# Hindsight Memory\nActive.",
-        memory_prefetch_raw="- remembered fact",
-        memory_context_block="<memory-context>\nremembered\n</memory-context>",
-        api_user_message="What do I prefer?\n\n<memory-context>\nremembered\n</memory-context>",
-        api_messages=[
-            {"role": "system", "content": "system prompt"},
-            {"role": "user", "content": "What do I prefer?"},
-        ],
-        metadata={"provider": "openai", "memory_providers": ["builtin", "hindsight"]},
-    )
-
-    assert out == tmp_path / "session-1" / "call_001"
-    assert (out / "system_prompt.txt").read_text(encoding="utf-8") == "system prompt"
-    assert (out / "memory_provider_system_prompt.txt").read_text(encoding="utf-8").startswith("# Hindsight Memory")
-    assert (out / "memory_prefetch_raw.txt").read_text(encoding="utf-8") == "- remembered fact"
-    assert "<memory-context>" in (out / "memory_context_block.txt").read_text(encoding="utf-8")
-    assert "What do I prefer?" in (out / "api_user_message.txt").read_text(encoding="utf-8")
-    assert (out / "api_messages.json").read_text(encoding="utf-8").startswith("[")
-    assert '"hindsight"' in (out / "metadata.json").read_text(encoding="utf-8")

tests/plugins/memory/test_hindsight_provider.py

@@ -596,26 +596,3 @@ class TestAvailability:
         monkeypatch.setenv("HINDSIGHT_MODE", "local")
         p = HindsightMemoryProvider()
         assert p.is_available()
-
-    def test_local_embedded_profile_defaults_to_agent_identity(self, tmp_path, monkeypatch):
-        config_path = tmp_path / "hindsight" / "config.json"
-        config_path.parent.mkdir(parents=True, exist_ok=True)
-        config_path.write_text(json.dumps({
-            "mode": "local_embedded",
-            "llm_provider": "ollama",
-            "llm_model": "gemma3:12b",
-        }))
-        monkeypatch.setattr(
-            "plugins.memory.hindsight.get_hermes_home",
-            lambda: tmp_path,
-        )
-
-        p = HindsightMemoryProvider()
-        p.initialize(
-            session_id="test-session",
-            hermes_home=str(tmp_path / "profiles" / "atlas-hindsight"),
-            platform="cli",
-            agent_identity="atlas-hindsight",
-        )
-
-        assert p._config["profile"] == "atlas-hindsight"

tests/plugins/memory/test_holographic_observations.py

@@ -0,0 +1,96 @@
+import json
+
+import pytest
+
+from plugins.memory.holographic import HolographicMemoryProvider
+from plugins.memory.holographic.store import MemoryStore
+
+
+@pytest.fixture()
+def store(tmp_path):
+    db_path = tmp_path / "memory.db"
+    s = MemoryStore(db_path=str(db_path), default_trust=0.5)
+    yield s
+    s.close()
+
+
+@pytest.fixture()
+def provider(tmp_path):
+    p = HolographicMemoryProvider(
+        config={
+            "db_path": str(tmp_path / "memory.db"),
+            "default_trust": 0.5,
+        }
+    )
+    p.initialize(session_id="test-session")
+    yield p
+    if p._store:
+        p._store.close()
+
+
+class TestObservationSynthesis:
+    def test_observe_action_persists_observation_with_evidence_links(self, provider):
+        fact_ids = [
+            provider._store.add_fact('User prefers concise status updates', category='user_pref'),
+            provider._store.add_fact('User wants result-only replies with no fluff', category='user_pref'),
+        ]
+
+        result = json.loads(
+            provider.handle_tool_call(
+                'fact_store',
+                {
+                    'action': 'observe',
+                    'query': 'What communication style does the user prefer?',
+                    'limit': 5,
+                },
+            )
+        )
+
+        assert result['count'] == 1
+        observation = result['observations'][0]
+        assert observation['observation_type'] == 'recurring_preference'
+        assert observation['confidence'] >= 0.6
+        assert sorted(item['fact_id'] for item in observation['evidence']) == sorted(fact_ids)
+
+        stored = provider._store.list_observations(limit=10)
+        assert len(stored) == 1
+        assert stored[0]['observation_type'] == 'recurring_preference'
+        assert stored[0]['evidence_count'] == 2
+        assert len(provider._store.list_facts(limit=10)) == 2
+
+    def test_observe_action_synthesizes_three_observation_types(self, provider):
+        provider._store.add_fact('User prefers concise updates', category='user_pref')
+        provider._store.add_fact('User wants result-only communication', category='user_pref')
+        provider._store.add_fact('Project is moving to a local-first deployment model', category='project')
+        provider._store.add_fact('Project direction stays Gitea-first for issue and PR flow', category='project')
+        provider._store.add_fact('Operator always commits early before moving on', category='general')
+        provider._store.add_fact('Operator pushes a PR immediately after each meaningful fix', category='general')
+
+        result = json.loads(provider.handle_tool_call('fact_store', {'action': 'observe', 'limit': 10}))
+        types = {item['observation_type'] for item in result['observations']}
+
+        assert {'recurring_preference', 'stable_direction', 'behavioral_pattern'} <= types
+
+    def test_single_fact_does_not_create_overconfident_observation(self, provider):
+        provider._store.add_fact('User prefers concise updates', category='user_pref')
+
+        result = json.loads(
+            provider.handle_tool_call(
+                'fact_store',
+                {'action': 'observe', 'query': 'What does the user prefer?', 'limit': 5},
+            )
+        )
+
+        assert result['count'] == 0
+        assert provider._store.list_observations(limit=10) == []
+
+    def test_prefetch_surfaces_observations_as_separate_layer(self, provider):
+        provider._store.add_fact('User prefers concise updates', category='user_pref')
+        provider._store.add_fact('User wants result-only communication', category='user_pref')
+
+        prefetch = provider.prefetch('What communication style does the user prefer?')
+
+        assert '## Holographic Observations' in prefetch
+        assert '## Holographic Memory' in prefetch
+        assert 'recurring_preference' in prefetch
+        assert 'evidence' in prefetch.lower()

tests/run_agent/test_run_agent_context_snapshots.py

@@ -1,94 +0,0 @@
-from pathlib import Path
-from types import SimpleNamespace
-from unittest.mock import MagicMock, patch
-import importlib
-import sys
-import types
-
-
-
-def _make_tool_defs(*names: str) -> list:
-    return [
-        {
-            "type": "function",
-            "function": {
-                "name": n,
-                "description": f"{n} tool",
-                "parameters": {"type": "object", "properties": {}},
-            },
-        }
-        for n in names
-    ]
-
-
-
-def _mock_response(content="Done", finish_reason="stop"):
-    msg = SimpleNamespace(content=content, tool_calls=None)
-    choice = SimpleNamespace(message=msg, finish_reason=finish_reason)
-    return SimpleNamespace(choices=[choice], usage=SimpleNamespace(prompt_tokens=1, completion_tokens=1, total_tokens=2))
-
-
-
-def _load_ai_agent():
-    sys.modules.setdefault("agent.auxiliary_client", types.SimpleNamespace(call_llm=lambda *a, **k: ""))
-    run_agent = importlib.import_module("run_agent")
-    return run_agent.AIAgent
-
-
-
-def test_run_conversation_writes_context_snapshot_artifacts(tmp_path):
-    AIAgent = _load_ai_agent()
-
-    class _FakePrivacyFilter:
-        def __init__(self):
-            self.last_report = None
-
-        def sanitize_messages(self, messages):
-            return list(messages)
-
-    with (
-        patch("run_agent.get_tool_definitions", return_value=_make_tool_defs("web_search")),
-        patch("run_agent.check_toolset_requirements", return_value={}),
-        patch("run_agent.OpenAI"),
-        patch("hermes_cli.plugins.invoke_hook", return_value=[]),
-        patch.dict(sys.modules, {"agent.privacy_filter": types.SimpleNamespace(PrivacyFilter=_FakePrivacyFilter)}),
-    ):
-        agent = AIAgent(
-            api_key="test-key-1234567890",
-            base_url="https://example.com/v1",
-            quiet_mode=True,
-            skip_context_files=True,
-            skip_memory=True,
-            context_snapshots_enabled=True,
-            context_snapshots_dir=str(tmp_path),
-        )
-
-    agent.client = MagicMock()
-    agent.client.chat.completions.create.return_value = _mock_response(content="Done")
-    agent._build_system_prompt = MagicMock(return_value="Core system prompt")
-    agent._memory_manager = MagicMock()
-    agent._memory_manager.prefetch_all.return_value = "- remembered preference"
-    agent._memory_manager.build_system_prompt.return_value = "# Hindsight Memory\nActive."
-    agent._memory_manager.providers = [
-        SimpleNamespace(name="builtin"),
-        SimpleNamespace(name="hindsight"),
-    ]
-
-    result = agent.run_conversation("What do I prefer?")
-
-    assert result["final_response"] == "Done"
-
-    call_dir = tmp_path / agent.session_id / "call_001"
-    assert call_dir.exists()
-    assert (call_dir / "system_prompt.txt").read_text(encoding="utf-8") == "Core system prompt"
-    assert (call_dir / "memory_provider_system_prompt.txt").read_text(encoding="utf-8").startswith("# Hindsight Memory")
-    assert (call_dir / "memory_prefetch_raw.txt").read_text(encoding="utf-8") == "- remembered preference"
-    assert "<memory-context>" in (call_dir / "memory_context_block.txt").read_text(encoding="utf-8")
-    api_user_message = (call_dir / "api_user_message.txt").read_text(encoding="utf-8")
-    assert "What do I prefer?" in api_user_message
-    assert "remembered preference" in api_user_message
-    api_messages = (call_dir / "api_messages.json").read_text(encoding="utf-8")
-    assert '"role": "system"' in api_messages
-    assert '"role": "user"' in api_messages
-    metadata = (call_dir / "metadata.json").read_text(encoding="utf-8")
-    assert '"hindsight"' in metadata

tests/test_hindsight_local_eval_doc.py

@@ -1,22 +0,0 @@
-from pathlib import Path
-
-
-ROOT = Path(__file__).resolve().parents[1]
-DOC = ROOT / "docs" / "hindsight-local-eval.md"
-
-
-def test_hindsight_local_eval_doc_exists_and_covers_switching():
-    assert DOC.exists(), "missing Hindsight local eval doc"
-    text = DOC.read_text(encoding="utf-8")
-    for snippet in (
-        "atlas-baseline",
-        "atlas-mempalace",
-        "atlas-hindsight",
-        "HERMES_HOME",
-        "HERMES_CONTEXT_SNAPSHOTS",
-        "memory_prefetch_raw.txt",
-        "api_user_message.txt",
-        "local_embedded",
-        "hindsight/config.json",
-    ):
-        assert snippet in text