feat: v0.7.0 — search quality, security hardening, CI pipeline

- Score-gap dedup, daily penalty, canonical question, raw cosine exposure
- Confidence field + content_hash in frontmatter
- CORS, rate limiting, request size limits, stack trace sanitization
- MCP audit log (structured JSONL tool call logging)
- GitHub Actions CI pipeline (unit tests + security scan)
- Integration test suite (14 API roundtrip tests, 175 total)
- Multi-platform MCP setup guide (Claude Code, Cursor, VS Code, Zed, Windsurf)
- PyPI-ready pyproject.toml
- timeline merged into history (17 tools)
- Migration module removed

Author: Paul Kyle

claude-plugin/.claude-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "name": "palinode",
   "description": "Persistent memory for AI agents. Git-versioned markdown files as source of truth, hybrid SQLite-vec + FTS5 search, deterministic compaction. Every line in your agent's brain has a git blame.",
-  "version": "0.6.2",
+  "version": "0.7.0",
   "author": {
     "name": "Paul Kyle",
     "url": "https://github.com/Paul-Kyle"

claude-plugin/README.md

@@ -161,7 +161,7 @@ Run `palinode_status` and check `total_files` and `fts_chunks`. If both are 0, t
 ## Learn more
 
 - [Main repository](https://github.com/Paul-Kyle/palinode)
-- [CHANGELOG](https://github.com/Paul-Kyle/palinode/blob/main/docs/CHANGELOG.md) for what's in v0.6.2
+- [CHANGELOG](https://github.com/Paul-Kyle/palinode/blob/main/docs/CHANGELOG.md) for what's in v0.7.0
 - [Compaction demo](https://github.com/Paul-Kyle/palinode/tree/main/examples/compaction-demo) — walkthrough of a memory file across three consolidation passes with blame + diff output
 - [ADR-001: Tools Over Pipeline](https://github.com/Paul-Kyle/palinode/blob/main/ADR-001-tools-over-pipeline.md) — why the executor is deterministic
 

docs/CHANGELOG.md

@@ -2,94 +2,70 @@
 
 All notable changes to Palinode. Format follows [Keep a Changelog](https://keepachangelog.com/en/1.1.0/).
 
-## [0.6.2] — 2026-04-12
+## [0.7.0] — 2026-04-12
 
 ### Added
 
-**Multi-platform MCP setup guide**
-- `docs/MCP-SETUP.md` — setup instructions for Claude Code, Cursor, VS Code (Continue + Cline), Zed, and Windsurf
-- Replaces the old Claude Code-only setup doc
+**Search quality**
+- **Score-gap dedup** (#91) — additional chunks from the same file are kept only if within `dedup_score_gap` (default 0.2) of the file's best score. Reduces noise from multi-section files dominating results.
+- **G1 context boost fix** (#92) — `store.search()` now accepts `context_entities` for ADR-008 ambient context boost. Previously the boost only fired through `search_hybrid`.
+- **Raw cosine exposure** (#94) — search results include a `raw_score` field with the original cosine similarity before RRF normalization.
+- **Daily penalty** (#93) — `daily/` files receive `score * daily_penalty` (default 0.3) to prevent daily notes from dominating search results. New `include_daily` parameter opts out of the penalty. Exposed as an MCP tool parameter.
+- **Canonical question frontmatter** (#83) — `canonical_question` frontmatter field (string or list) is prepended as `"Q: ..."` to the first chunk before embedding, anchoring each memory to the question it answers.
+- **Confidence field + content_hash in frontmatter** (#113, #114) — new `confidence` field for memory files, full SHA-256 content hash stored in frontmatter for integrity verification.
 
-**Agent plugin restored**
-- `plugin/` re-added after scrubbing internal references
-- Hook migrated from `before_agent_start` to `before_prompt_build`
-- Fixed `cfg.apiUrl` → `cfg.palinodeApiUrl` runtime bug
+**Security**
+- CORS, rate limiting, request size limits, stack trace sanitization for the API server
+- MCP audit log — structured JSONL tool call logging (#116)
+
+**CI/CD**
+- GitHub Actions pipeline — unit tests + security scan (#121)
+
+**Testing**
+- Integration test suite — 14 API roundtrip tests (#120)
+- 175 tests passing (up from 149)
+
+**Documentation**
+- Multi-platform MCP setup guide (`docs/MCP-SETUP.md`) — Claude Code, Cursor, VS Code, Zed, Windsurf
+- PyPI-ready pyproject.toml with metadata and classifiers
 
 ### Changed
 
-- `palinode_timeline` merged into `palinode_history` — one tool with `--follow`, diff stats, structured JSON return, and `limit` parameter (closes #1)
+- `palinode_timeline` merged into `palinode_history` — one tool with `--follow`, diff stats, structured JSON return, and `limit` parameter
 - Tool count: 18 → 17 (timeline/history consolidated)
-- Claude Code plugin manifest updated to v0.6.2
+- README repositioned as memory substrate, not just persistent memory
 
 ### Removed
 
+- Migration endpoints and CLI commands (`palinode.migration` module removed)
 - `docs/claude-code-setup.md` — replaced by `docs/MCP-SETUP.md`
 
 ---
 
-## [0.6.1] — 2026-04-12
-
-### Added
-
-**RETRACT operation**
-- New executor operation: `RETRACT` — marks a memory fact as wrong with a visible tombstone
-- Strikethrough formatting with `[RETRACTED date — reason]` annotation
-- Fact ID preserved (not deleted) so readers know what was retracted and why
-- History file records retraction provenance
-- Compaction and update prompts updated with RETRACT guidance
-- 4 new tests (121 total)
-
-Maps to IETF Knowledge Unit `retract` lifecycle state — see Paul-Kyle/palinode#17 for the interop discussion.
-
----
-
 ## [0.6.0] — 2026-04-11
 
 ### Added
 
 **Write-time contradiction check (ADR-004)**
-- When saving a memory, the system now checks for contradictions against existing files in the same entity scope
-- Contradiction candidates are surfaced before the save completes, with configurable thresholds
-- Background worker runs via asyncio queue (API) or disk-backed marker files (CLI/plugin)
+- When saving a memory, the system checks for contradictions against existing files in the same entity scope
+- Contradiction candidates surfaced before the save completes, with configurable thresholds
 
 **Ambient context search (ADR-008)**
-- Search results are now boosted by project context inferred from the caller's working directory
+- Search results boosted by project context inferred from the caller's working directory
 - Resolution chain: `PALINODE_PROJECT` env var → config project map → CWD auto-detect
-- Existing RRF hybrid search pipeline extended with a context scoring channel
+
+**RETRACT operation**
+- New executor operation: `RETRACT` — marks a memory fact as wrong with a visible tombstone
+- Strikethrough formatting with `[RETRACTED date — reason]` annotation
+- Fact ID preserved so readers know what was retracted and why
 
 **Claude Code plugin scaffold**
 - `claude-plugin/` directory with plugin manifest for Claude Code marketplace submission
 
 **Claude Code skills**
 - `palinode-claude-code` — MCP setup and usage for Claude Code sessions
-- `palinode-memory` — general memory operations skill
 - `palinode-session` — automatic session lifecycle memory capture
 
-**Architecture Decision Records**
-- ADR-004: Event-driven consolidation (write-time contradiction check)
-- ADR-005: Debounced reflection executor
-- ADR-006: On-read reconsolidation
-- ADR-007: Access metadata and decay
-- ADR-008: Ambient context search
-
-**Documentation**
-- WHY-LOCAL-MEMORY.md — positioning document for local-first memory
-- Research paper: Memory Compaction and Augmented Recall for Persistent AI Agents
-- PRD (product requirements document)
-
-**Tests**
-- `test_write_time.py` — write-time contradiction check coverage
-- `test_context.py` — ambient context search and project boosting
-
-### Changed
-- MCP server uses Streamable HTTP transport (renamed from SSE entry point)
-- Search API accepts optional `context` parameter for entity-scoped boosting
-- Consolidation cron scheduling improvements
-
-### Removed
-- `palinode/migration/` — internal migration tooling removed
-- `plugin/` — old OpenClaw plugin (replaced by `claude-plugin/` scaffold)
-
 ---
 
 ## [0.5.0] — 2026-04-10
@@ -105,21 +81,21 @@ First tagged release. Persistent memory for AI agents with git-versioned markdow
 - File watcher daemon with debounced reindex and fault isolation
 
 **Consolidation and compaction**
-- Deterministic executor applying `KEEP` / `UPDATE` / `MERGE` / `SUPERSEDE` / `ARCHIVE` operations proposed by an LLM (see [ADR-001](../ADR-001-tools-over-pipeline.md))
+- Deterministic executor applying `KEEP` / `UPDATE` / `MERGE` / `SUPERSEDE` / `ARCHIVE` operations proposed by an LLM
 - Weekly full-corpus consolidation with configurable LLM backend
 - Nightly lightweight consolidation pass (`--nightly` flag) bounded to `UPDATE`/`SUPERSEDE` for safer incremental updates
 - Model fallback chains — primary → fallback → fallback on timeout or HTTP error
 - Prompt versioning system — extraction/compaction prompts stored as memory files with `active: true` frontmatter
 
 **Interfaces (all four expose the same capabilities)**
-- **MCP server** — Streamable HTTP transport (also supports stdio) with 18 tools. Stateless HTTP client, point it at any Palinode API server
+- **MCP server** — Streamable HTTP transport (also supports stdio). Stateless HTTP client, point it at any Palinode API server
 - **REST API** — FastAPI on port 6340, 20+ endpoints covering search, save, diff, triggers, history, blame, rollback, consolidation, session-end, lint
 - **CLI** — 26 commands wrapping the REST API via Click. TTY-aware (human output interactive, JSON when piped). Remote access via `PALINODE_API` env var
 - **Plugin** — OpenClaw lifecycle hooks for agent frameworks with inject/extract patterns
 
 **New MCP tools in this release**
 - `palinode_lint` — scan memory for orphaned files, stale active files, missing frontmatter, potential contradictions
-- `palinode_blame` / `palinode_history` / `palinode_timeline` / `palinode_rollback` — git-backed provenance tools
+- `palinode_blame` / `palinode_history` / `palinode_rollback` — git-backed provenance tools
 - `palinode_trigger` — register prospective triggers that inject memory files when matching context is detected
 - `palinode_prompt` — list, read, and activate versioned LLM prompt files
 
@@ -134,13 +110,12 @@ First tagged release. Persistent memory for AI agents with git-versioned markdow
 - Exclude-paths list prevents search results from surfacing files in `.secrets`, `credentials`, etc.
 
 **Documentation**
-- [ADR-001: Tools Over Pipeline](../ADR-001-tools-over-pipeline.md) — why the executor is deterministic
 - Remote MCP setup guides for Claude Code, Claude Desktop, Cursor, Zed
 - Example memory files (`examples/people/`, `examples/projects/`, `examples/decisions/`, `examples/insights/`)
 - Compaction walkthrough (`examples/compaction-demo/`) — a memory file across 3 passes with blame + diff output
 
 **Tests**
-- 92 tests covering parser, store, executor, API, CLI, migration, and hybrid search
+- 92 tests covering parser, store, executor, API, CLI, and hybrid search
 
 ### Changed
 - All inference is local by default. Cloud API keys (Gemini, OpenAI) are opt-in via environment variables
@@ -151,7 +126,6 @@ First tagged release. Persistent memory for AI agents with git-versioned markdow
 ### Fixed
 - Watcher no longer crashes the API server if the memory directory is temporarily unavailable
 - CLI display keys match API response keys across all commands
-- Migration tool correctly handles frontmatter with embedded colons
 
 ### Removed
 - Deprecated SSE MCP transport (replaced by Streamable HTTP per canonical MCP SDK pattern)

docs/HOW-MEMORY-WORKS.md

@@ -297,7 +297,7 @@ Every memory change is a git commit. This enables:
 | --- | --- | --- |
 | `palinode_diff` | What changed recently? | "Show me changes this week" |
 | `palinode_blame` | When was this fact recorded? | "When did Alice mention async?" |
-| `palinode_timeline` | How has this file evolved? | "Show My App's history" |
+| `palinode_history` | How has this file evolved? | "Show My App's history" |
 | `palinode_rollback` | Undo a bad change | "Revert last consolidation" |
 | `palinode_push` | Sync to GitHub | "Backup my memory" |
 
@@ -416,7 +416,7 @@ Other memory systems are opaque databases. You can query them but you can't ask:
 
 - "When did I first learn that Alice prefers async?" → `palinode_blame`
 - "What changed about the My App project this week?" → `palinode_diff`
-- "Show me every update to my infrastructure notes" → `palinode_timeline`
+- "Show me every update to my infrastructure notes" → `palinode_history`
 - "The last consolidation was bad, undo it" → `palinode_rollback`
 
 These aren't add-on features. They're consequences of the architectural decision to use files + git as the source of truth. The audit trail is free.

docs/INSTALL-CLAUDE-CODE.md

@@ -284,7 +284,7 @@ Search palinode for "recent project decisions"
 | `palinode_consolidate` | Run or preview the weekly compaction job |
 | `palinode_diff` | See what changed in memory recently (git diff) |
 | `palinode_blame` | Who/when each line was written (git blame) |
-| `palinode_timeline` | Chronological changes to a memory file |
+| `palinode_history` | Git history with diff stats and rename tracking |
 | `palinode_rollback` | Revert a file to a previous state |
 | `palinode_push` | Push memory changes to remote git |
 | `palinode_trigger` | Register a prospective recall trigger |

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "palinode"
-version = "0.6.2"
+version = "0.7.0"
 description = "Persistent memory that makes AI agents smarter over time."
 authors = [
   {name = "Paul Kyle"}

skill/palinode-claude-code/references/setup.md

@@ -156,7 +156,7 @@ Search palinode for "recent project decisions"
 | `palinode_consolidate` | Run or preview the weekly compaction job |
 | `palinode_diff` | See what changed in a memory file (git diff) |
 | `palinode_blame` | Who/when each section was written |
-| `palinode_timeline` | Activity over time |
+| `palinode_history` | Git history with diff stats and rename tracking |
 | `palinode_rollback` | Revert a file to a previous state |
 | `palinode_push` | Push memory changes to remote git |
 | `palinode_trigger` | Register a prospective recall intention |

skill/palinode-memory/SKILL.md

@@ -86,7 +86,7 @@ Triggers fire when the user's message semantically matches `description`. Good f
 Use these when the user asks about memory history:
 - `palinode_diff` — what changed in the last N commits for a file
 - `palinode_blame` — who/when each section was written
-- `palinode_timeline` — activity over time
+- `palinode_history` — git history with diff stats and rename tracking
 - `palinode_rollback` — revert a file to a previous state
 
 ## Consolidation

tests/test_context.py

@@ -11,21 +11,21 @@ def test_search_hybrid_context_boost():
         with patch("palinode.core.store.search_fts") as mock_fts:
             with patch("palinode.core.store.get_db"):
                 with patch("palinode.core.store.get_entity_files") as mock_entities:
-                    # Two results: palinode file and other-project file, initially ranked equally
+                    # Two results: palinode file and acme-project file, initially ranked equally
                     mock_vec.return_value = [
                         {"file_path": "/mem/projects/palinode-adr.md", "content": "ADR-004", "score": 0.85},
-                        {"file_path": "/mem/projects/kmd-adr.md", "content": "ADR-052", "score": 0.86},
+                        {"file_path": "/mem/projects/acme-adr.md", "content": "ADR-052", "score": 0.86},
                     ]
                     mock_fts.return_value = [
-                        {"file_path": "/mem/projects/kmd-adr.md", "content": "ADR-052", "score": 0.7},
+                        {"file_path": "/mem/projects/acme-adr.md", "content": "ADR-052", "score": 0.7},
                         {"file_path": "/mem/projects/palinode-adr.md", "content": "ADR-004", "score": 0.6},
                     ]
                     # Entity lookup: project/palinode maps to the palinode file
                     mock_entities.return_value = [
                         {"file_path": "/mem/projects/palinode-adr.md", "category": "projects", "last_seen": "2026-04-12"}
                     ]
 
-                    # Without context: other-project should rank first (higher combined score)
+                    # Without context: acme-project should rank first (higher combined score)
                     results_no_ctx = store.search_hybrid(
                         "ADR-004", query_embedding=[0.0]*1024, top_k=2, threshold=0.0,
                         context_entities=None,
@@ -107,6 +107,94 @@ def test_search_hybrid_boost_factor():
         config.context.boost = original
 
 
+def test_search_vector_context_boost():
+    """Non-hybrid search should also apply context boost (#92)."""
+    with patch("palinode.core.store.get_db") as mock_db:
+        with patch("palinode.core.store.get_entity_files") as mock_entities:
+            # Simulate two vector results: acme ranks higher by raw cosine
+            mock_cursor = MagicMock()
+            mock_cursor.fetchall.return_value = [
+                {"id": 1, "file_path": "/mem/projects/acme-adr.md", "section_id": "root",
+                 "content": "ADR-052", "category": "projects", "metadata": "{}",
+                 "created_at": "2026-04-12", "distance": 0.3},
+                {"id": 2, "file_path": "/mem/projects/palinode-adr.md", "section_id": "root",
+                 "content": "ADR-004", "category": "projects", "metadata": "{}",
+                 "created_at": "2026-04-12", "distance": 0.35},
+            ]
+            mock_db.return_value.cursor.return_value = mock_cursor
+            mock_db.return_value.close = MagicMock()
+
+            mock_entities.return_value = [
+                {"file_path": "/mem/projects/palinode-adr.md", "category": "projects", "last_seen": "2026-04-12"}
+            ]
+
+            # Without context: acme first (lower distance = higher score)
+            results_no_ctx = store.search(
+                query_embedding=[0.0]*1024, top_k=2, threshold=0.0,
+                context_entities=None,
+            )
+            assert len(results_no_ctx) == 2
+            assert "acme" in results_no_ctx[0]["file_path"]
+
+            # With context: palinode should be boosted to first
+            results_ctx = store.search(
+                query_embedding=[0.0]*1024, top_k=2, threshold=0.0,
+                context_entities=["project/palinode"],
+            )
+            assert len(results_ctx) == 2
+            assert "palinode" in results_ctx[0]["file_path"]
+
+
+def test_search_vector_context_disabled_no_boost():
+    """Non-hybrid search should not boost when context.enabled is False (#92)."""
+    original = config.context.enabled
+    try:
+        config.context.enabled = False
+        with patch("palinode.core.store.get_db") as mock_db:
+            with patch("palinode.core.store.get_entity_files") as mock_entities:
+                mock_cursor = MagicMock()
+                mock_cursor.fetchall.return_value = [
+                    {"id": 1, "file_path": "a.md", "section_id": "root",
+                     "content": "text", "category": "projects", "metadata": "{}",
+                     "created_at": "2026-04-12", "distance": 0.3},
+                ]
+                mock_db.return_value.cursor.return_value = mock_cursor
+                mock_db.return_value.close = MagicMock()
+
+                store.search(
+                    query_embedding=[0.0]*1024, top_k=2, threshold=0.0,
+                    context_entities=["project/palinode"],
+                )
+                mock_entities.assert_not_called()
+    finally:
+        config.context.enabled = original
+
+
+def test_search_vector_boost_factor_one_noop():
+    """Non-hybrid search: boost=1.0 should be a no-op (#92)."""
+    original = config.context.boost
+    try:
+        config.context.boost = 1.0
+        with patch("palinode.core.store.get_db") as mock_db:
+            with patch("palinode.core.store.get_entity_files") as mock_entities:
+                mock_cursor = MagicMock()
+                mock_cursor.fetchall.return_value = [
+                    {"id": 1, "file_path": "a.md", "section_id": "root",
+                     "content": "text", "category": "projects", "metadata": "{}",
+                     "created_at": "2026-04-12", "distance": 0.3},
+                ]
+                mock_db.return_value.cursor.return_value = mock_cursor
+                mock_db.return_value.close = MagicMock()
+
+                store.search(
+                    query_embedding=[0.0]*1024, top_k=2, threshold=0.0,
+                    context_entities=["project/palinode"],
+                )
+                mock_entities.assert_not_called()
+    finally:
+        config.context.boost = original
+
+
 def test_context_config_defaults():
     """ContextConfig should have sane defaults."""
     assert config.context.enabled is True