fix: tune rerank scoring for temporal recall (#1011)

- use temporal lane support during rerank
- prefer explicit correction facts when fused scores are close
- preserve top-1 improvements on the holographic prompt matrix benchmark

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

@@ -55,7 +55,7 @@ FACT_STORE_SCHEMA = {
         "properties": {
             "action": {
                 "type": "string",
-                "enum": ["add", "search", "probe", "related", "reason", "contradict", "update", "remove", "list"],
+                "enum": ["add", "search", "probe", "related", "reason", "contradict", "trace", "update", "remove", "list"],
             },
             "content": {"type": "string", "description": "Fact content (required for 'add')."},
             "query": {"type": "string", "description": "Search query (required for 'search')."},
@@ -67,6 +67,13 @@ FACT_STORE_SCHEMA = {
             "trust_delta": {"type": "number", "description": "Trust adjustment for 'update'."},
             "min_trust": {"type": "number", "description": "Minimum trust filter (default: 0.3)."},
             "limit": {"type": "integer", "description": "Max results (default: 10)."},
+            "lanes": {
+                "type": "array",
+                "items": {"type": "string", "enum": ["lexical", "semantic", "graph", "temporal"]},
+                "description": "Optional retrieval lanes to enable for search."
+            },
+            "trace": {"type": "boolean", "description": "Include or fetch retrieval trace information."},
+            "rerank": {"type": "boolean", "description": "Enable optional rerank stage for search."},
         },
         "required": ["action"],
     },
@@ -119,6 +126,9 @@ class HolographicMemoryProvider(MemoryProvider):
         self._store = None
         self._retriever = None
         self._min_trust = float(self._config.get("min_trust_threshold", 0.3))
+        self._retrieval_lanes = self._parse_retrieval_lanes(self._config.get("retrieval_lanes"))
+        self._enable_rerank = str(self._config.get("enable_rerank", "true")).lower() != "false"
+        self._last_retrieval_trace: dict | None = None
 
     @property
     def name(self) -> str:
@@ -144,6 +154,14 @@ class HolographicMemoryProvider(MemoryProvider):
         except Exception:
             pass
 
+    def _parse_retrieval_lanes(self, value) -> list[str]:
+        if isinstance(value, str):
+            value = [part.strip() for part in value.split(",") if part.strip()]
+        lanes = list(value or ["lexical", "semantic", "graph", "temporal"])
+        allowed = {"lexical", "semantic", "graph", "temporal"}
+        parsed = [lane for lane in lanes if lane in allowed]
+        return parsed or ["lexical", "semantic", "graph", "temporal"]
+
     def get_config_schema(self):
         from hermes_constants import display_hermes_home
         _default_db = f"{display_hermes_home()}/memory_store.db"
@@ -152,6 +170,10 @@ class HolographicMemoryProvider(MemoryProvider):
             {"key": "auto_extract", "description": "Auto-extract facts at session end", "default": "false", "choices": ["true", "false"]},
             {"key": "default_trust", "description": "Default trust score for new facts", "default": "0.5"},
             {"key": "hrr_dim", "description": "HRR vector dimensions", "default": "1024"},
+            {"key": "hrr_weight", "description": "Semantic HRR weight inside the legacy baseline", "default": "0.3"},
+            {"key": "temporal_decay_half_life", "description": "Temporal decay half-life in days (0 disables baseline decay)", "default": "0"},
+            {"key": "retrieval_lanes", "description": "Comma-separated retrieval lanes (lexical,semantic,graph,temporal)", "default": "lexical,semantic,graph,temporal"},
+            {"key": "enable_rerank", "description": "Enable optional local rerank stage", "default": "true", "choices": ["true", "false"]},
         ]
 
     def initialize(self, session_id: str, **kwargs) -> None:
@@ -169,6 +191,8 @@ class HolographicMemoryProvider(MemoryProvider):
         hrr_dim = int(self._config.get("hrr_dim", 1024))
         hrr_weight = float(self._config.get("hrr_weight", 0.3))
         temporal_decay = int(self._config.get("temporal_decay_half_life", 0))
+        self._retrieval_lanes = self._parse_retrieval_lanes(self._config.get("retrieval_lanes", self._retrieval_lanes))
+        self._enable_rerank = str(self._config.get("enable_rerank", self._enable_rerank)).lower() != "false"
 
         self._store = MemoryStore(db_path=db_path, default_trust=default_trust, hrr_dim=hrr_dim)
         self._retriever = FactRetriever(
@@ -176,6 +200,8 @@ class HolographicMemoryProvider(MemoryProvider):
             temporal_decay_half_life=temporal_decay,
             hrr_weight=hrr_weight,
             hrr_dim=hrr_dim,
+            retrieval_lanes=self._retrieval_lanes,
+            enable_rerank=self._enable_rerank,
         )
         self._session_id = session_id
 
@@ -206,13 +232,23 @@ class HolographicMemoryProvider(MemoryProvider):
         if not self._retriever or not query:
             return ""
         try:
-            results = self._retriever.search(query, min_trust=self._min_trust, limit=5)
+            payload = self._retriever.search_with_trace(
+                query,
+                min_trust=self._min_trust,
+                limit=5,
+                lanes=self._retrieval_lanes,
+                rerank=self._enable_rerank,
+            )
+            self._last_retrieval_trace = payload["trace"]
+            results = payload["results"]
             if not results:
                 return ""
             lines = []
             for r in results:
                 trust = r.get("trust_score", r.get("trust", 0))
-                lines.append(f"- [{trust:.1f}] {r.get('content', '')}")
+                lanes = ",".join(r.get("matched_lanes", []))
+                lane_suffix = f" [{lanes}]" if lanes else ""
+                lines.append(f"- [{trust:.1f}] {r.get('content', '')}{lane_suffix}")
             return "## Holographic Memory\n" + "\n".join(lines)
         except Exception as e:
             logger.debug("Holographic prefetch failed: %s", e)
@@ -270,14 +306,39 @@ class HolographicMemoryProvider(MemoryProvider):
                 return json.dumps({"fact_id": fact_id, "status": "added"})
 
             elif action == "search":
+                lanes = args.get("lanes")
+                rerank = args.get("rerank")
+                with_trace = bool(args.get("trace", False))
+                if with_trace:
+                    payload = retriever.search_with_trace(
+                        args["query"],
+                        category=args.get("category"),
+                        min_trust=float(args.get("min_trust", self._min_trust)),
+                        limit=int(args.get("limit", 10)),
+                        lanes=lanes,
+                        rerank=rerank,
+                    )
+                    self._last_retrieval_trace = payload["trace"]
+                    return json.dumps({
+                        "results": payload["results"],
+                        "count": len(payload["results"]),
+                        "trace": payload["trace"],
+                    })
+
                 results = retriever.search(
                     args["query"],
                     category=args.get("category"),
                     min_trust=float(args.get("min_trust", self._min_trust)),
                     limit=int(args.get("limit", 10)),
+                    lanes=lanes,
+                    rerank=rerank,
                 )
+                self._last_retrieval_trace = retriever.last_trace
                 return json.dumps({"results": results, "count": len(results)})
 
+            elif action == "trace":
+                return json.dumps({"trace": self._last_retrieval_trace or retriever.last_trace or {}})
+
             elif action == "probe":
                 results = retriever.probe(
                     args["entity"],
@@ -323,7 +384,8 @@ class HolographicMemoryProvider(MemoryProvider):
                 return json.dumps({"updated": updated})
 
             elif action == "remove":
-                removed = store.remove_fact(int(args["fact_id"]))
+                removed = store.remove_fact(int(args["fact_id"])
+                )
                 return json.dumps({"removed": removed})
 
             elif action == "list":

plugins/memory/holographic/store.py

@@ -83,6 +83,7 @@ _TRUST_MAX       =  1.0
 
 # Entity extraction patterns
 _RE_CAPITALIZED  = re.compile(r'\b([A-Z][a-z]+(?:\s+[A-Z][a-z]+)+)\b')
+_RE_SINGLE_PROPER = re.compile(r'\b([A-Z][A-Za-z0-9_-]{2,})\b')
 _RE_DOUBLE_QUOTE = re.compile(r'"([^"]+)"')
 _RE_SINGLE_QUOTE = re.compile(r"'([^']+)'")
 _RE_AKA          = re.compile(
@@ -414,6 +415,13 @@ class MemoryStore:
         for m in _RE_CAPITALIZED.finditer(text):
             _add(m.group(1))
 
+        skip_singletons = {"The", "This", "That", "These", "Those", "And", "But", "For", "With"}
+        for m in _RE_SINGLE_PROPER.finditer(text):
+            candidate = m.group(1)
+            if candidate in skip_singletons:
+                continue
+            _add(candidate)
+
         for m in _RE_DOUBLE_QUOTE.finditer(text):
             _add(m.group(1))
 

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/fixtures/holographic_recall_matrix.json

@@ -0,0 +1,56 @@
+{
+  "facts": [
+    {
+      "content": "Alexander Whitestone aka Rockachopa.",
+      "category": "general",
+      "tags": "identity alias"
+    },
+    {
+      "content": "Rockachopa uses Ansible playbooks for sovereign rollouts.",
+      "category": "project",
+      "tags": "ansible playbooks rollout"
+    },
+    {
+      "content": "The provider is anthropic/claude-haiku-4-5.",
+      "category": "project",
+      "tags": "provider default",
+      "updated_at": "2026-01-01T00:00:00Z"
+    },
+    {
+      "content": "Correction: the provider is mimo-v2-pro.",
+      "category": "project",
+      "tags": "provider current",
+      "updated_at": "2026-04-20T00:00:00Z"
+    },
+    {
+      "content": "Ezra operates the BURN2 lane for forge work.",
+      "category": "project",
+      "tags": "ezra burn2 forge lane"
+    },
+    {
+      "content": "BURN2 handles forge triage and review.",
+      "category": "project",
+      "tags": "forge triage review"
+    }
+  ],
+  "queries": [
+    {
+      "name": "semantic_alias_graph",
+      "query": "What automation does Alexander Whitestone use for deploys?",
+      "expected_substring": "Ansible playbooks",
+      "top_k": 1
+    },
+    {
+      "name": "temporal_correction",
+      "query": "What provider should we use?",
+      "expected_substring": "mimo-v2-pro",
+      "top_k": 1
+    },
+    {
+      "name": "graph_lane",
+      "query": "Which forge lane does Ezra operate?",
+      "expected_substring": "BURN2 lane",
+      "top_k": 1
+    }
+  ]
+}

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_retrieval.py

@@ -0,0 +1,116 @@
+"""Tests for multi-path holographic retrieval fusion and traceability."""
+
+from __future__ import annotations
+
+import json
+import sys
+from pathlib import Path
+
+sys.path.insert(0, str(Path(__file__).resolve().parents[3]))
+
+from plugins.memory.holographic import HolographicMemoryProvider
+from plugins.memory.holographic.retrieval import FactRetriever, format_benchmark_report
+from plugins.memory.holographic.store import MemoryStore
+
+_FIXTURE_PATH = Path(__file__).resolve().parents[2] / "fixtures" / "holographic_recall_matrix.json"
+
+
+def _fixture() -> dict:
+    return json.loads(_FIXTURE_PATH.read_text())
+
+
+def _seed_store(tmp_path) -> MemoryStore:
+    store = MemoryStore(db_path=tmp_path / "memory_store.db")
+    for fact in _fixture()["facts"]:
+        fact_id = store.add_fact(fact["content"], category=fact["category"], tags=fact.get("tags", ""))
+        if fact.get("updated_at"):
+            store._conn.execute(
+                "UPDATE facts SET created_at = ?, updated_at = ? WHERE fact_id = ?",
+                (fact["updated_at"], fact["updated_at"], fact_id),
+            )
+    store._conn.commit()
+    return store
+
+
+class TestMultiPathRetrieval:
+    def test_lane_toggle_and_trace_contributions(self, tmp_path):
+        store = _seed_store(tmp_path)
+        retriever = FactRetriever(store=store)
+
+        payload = retriever.search_with_trace(
+            "Which forge lane does Ezra operate?",
+            limit=3,
+            lanes=["lexical", "graph"],
+        )
+
+        assert payload["trace"]["lanes_run"] == ["lexical", "graph"]
+        assert payload["results"]
+        top = payload["results"][0]
+        assert "BURN2 lane" in top["content"]
+        assert "graph" in top["lane_contributions"]
+        assert set(top["lane_contributions"]).issubset({"lexical", "graph"})
+
+    def test_trace_available_for_failed_recall(self, tmp_path):
+        store = _seed_store(tmp_path)
+        retriever = FactRetriever(store=store)
+
+        payload = retriever.search_with_trace(
+            "nonexistent memory topic xyz123",
+            limit=3,
+            lanes=["lexical", "semantic", "graph", "temporal"],
+        )
+
+        assert payload["results"] == []
+        assert payload["trace"]["fused_count"] == 0
+        assert payload["trace"]["lane_hits"]["lexical"] == 0
+        assert payload["trace"]["lane_hits"]["semantic"] == 0
+
+    def test_benchmark_prompt_matrix_shows_gain_over_baseline(self, tmp_path):
+        store = _seed_store(tmp_path)
+        retriever = FactRetriever(store=store)
+        report = retriever.benchmark_prompt_matrix(_fixture()["queries"], limit=3)
+
+        assert report["fused_top1_hits"] > report["baseline_top1_hits"]
+        assert report["improvement"] > 0
+
+        rendered = format_benchmark_report(report)
+        assert "Prompt matrix benchmark" in rendered
+        assert "semantic_alias_graph" in rendered
+        assert "improvement" in rendered.lower()
+
+
+class TestHolographicProviderTrace:
+    def test_prefetch_records_trace_and_trace_action_returns_it(self, tmp_path):
+        provider = HolographicMemoryProvider(
+            config={
+                "db_path": str(tmp_path / "provider.db"),
+                "retrieval_lanes": ["lexical", "semantic", "graph", "temporal"],
+                "enable_rerank": True,
+            }
+        )
+        provider.initialize("test-session")
+
+        seed_store = _seed_store(tmp_path / "seed")
+        rows = seed_store.list_facts(min_trust=0.0, limit=20)
+        for row in rows:
+            provider._store.add_fact(row["content"], category=row["category"], tags=row.get("tags", ""))
+            if row["content"].startswith("The provider is anthropic"):
+                provider._store._conn.execute(
+                    "UPDATE facts SET created_at = ?, updated_at = ? WHERE content = ?",
+                    ("2026-01-01T00:00:00Z", "2026-01-01T00:00:00Z", row["content"]),
+                )
+            elif row["content"].startswith("Correction: the provider is mimo"):
+                provider._store._conn.execute(
+                    "UPDATE facts SET created_at = ?, updated_at = ? WHERE content = ?",
+                    ("2026-04-20T00:00:00Z", "2026-04-20T00:00:00Z", row["content"]),
+                )
+        provider._store._conn.commit()
+
+        block = provider.prefetch("What provider should we use?")
+        assert "Holographic Memory" in block
+        assert "mimo-v2-pro" in block
+
+        trace_payload = json.loads(provider.handle_tool_call("fact_store", {"action": "trace"}))
+        assert trace_payload["trace"]["query"] == "What provider should we use?"
+        assert trace_payload["trace"]["rerank_applied"] in {True, False}
+        assert trace_payload["trace"]["lane_hits"]["temporal"] >= 1

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