chore(release): v0.18.0 -> main

.github/workflows/cla-check.yml

@@ -4,10 +4,13 @@ on:
   pull_request:
     types: [opened, synchronize, reopened]
 
+# pull-requests:write is only effective for SAME-REPO PRs (bot comment).
+# Fork PRs run with a read-only GITHUB_TOKEN no matter what is declared here
+# — that is exactly why the old commit-status approach could never work on
+# forks (verified: PR #1221, run 34481899319, POST statuses → 403).
 permissions:
   pull-requests: write
   contents: read
-  statuses: write
 
 jobs:
   cla-check:
@@ -46,8 +49,16 @@ jobs:
         env:
           PR_AUTHOR: ${{ github.event.pull_request.user.login }}
 
-      - name: Comment on PR (unsigned only)
-        if: steps.check.outputs.signed != 'true'
+      # FORK-SAFE GATE: this job's own Actions check run carries the CLA gate.
+      # It is created by GitHub itself and is NOT subject to the fork token
+      # restriction: signed → job success; unsigned → annotated failure below.
+      #
+      # The PR comment is a courtesy for same-repo contributors only — forks
+      # cannot comment at all (also 403), so the step is conditioned and
+      # best-effort. It never affects the gate outcome.
+      - name: Comment on PR (same-repo, best effort)
+        if: steps.check.outputs.signed != 'true' && github.event.pull_request.head.repo.full_name == github.repository
+        continue-on-error: true
         uses: actions/github-script@v9
         with:
           script: |
@@ -64,6 +75,7 @@ jobs:
                   '- 📋 **Individual?** → [Sign CLA Individual](https://codecoradev.github.io/cla/?type=individual)',
                   '- 🏢 **Corporate?** → [Sign CLA Corporate](https://codecoradev.github.io/cla/?type=corporate)',
                   '',
+                  'After signing, this check turns green automatically (within ~1 hour, or ask a maintainer to re-run it).',
                   '---',
                   '<sub>By signing, you agree to the terms in [CLA_INDIVIDUAL.md](https://github.com/codecoradev/.github/blob/main/CLA_INDIVIDUAL.md) or [CLA_CORPORATE.md](https://github.com/codecoradev/.github/blob/main/CLA_CORPORATE.md).</sub>',
                 ].join('\n');
@@ -96,27 +108,9 @@ jobs:
               });
             }
 
-      - name: Set commit status
-        uses: actions/github-script@v9
-        with:
-          script: |
-            const signed = '${{ steps.check.outputs.signed }}' === 'true';
-            await github.rest.repos.createCommitStatus({
-              owner: context.repo.owner,
-              repo: context.repo.repo,
-              sha: '${{ github.event.pull_request.head.sha }}',
-              state: signed ? 'success' : 'failure',
-              context: 'CLA Check',
-              description: signed
-                ? '✅ CLA signed'
-                : '❌ CLA not signed — sign at https://codecoradev.github.io/cla',
-              target_url: signed
-                ? 'https://github.com/codecoradev/.github/blob/main/.cla/signatures.json'
-                : 'https://codecoradev.github.io/cla',
-            });
-
       - name: Fail if not signed
         if: steps.check.outputs.signed != 'true'
         run: |
+          echo "::error title=CLA not signed::Sign the CodeCoraDev CLA at https://codecoradev.github.io/cla — once the signature PR is merged, this check turns green automatically (re-run it or wait ~1 hour)."
           echo "❌ CLA not signed by ${{ github.event.pull_request.user.login }}"
           exit 1

.github/workflows/pr-checks.yml

@@ -9,6 +9,7 @@ jobs:
   branch-naming:
     name: Branch Naming
     runs-on: ubuntu-latest
+    if: "!contains(fromJSON('[\"dependabot[bot]\", \"renovate[bot]\", \"github-actions[bot]\", \"app/dependabot\", \"app/renovate\"]'), github.event.pull_request.user.login)"
     steps:
       - name: Check branch name
         env:
@@ -38,6 +39,7 @@ jobs:
   pr-template:
     name: PR Description
     runs-on: ubuntu-latest
+    if: "!contains(fromJSON('[\"dependabot[bot]\", \"renovate[bot]\", \"github-actions[bot]\", \"app/dependabot\", \"app/renovate\"]'), github.event.pull_request.user.login)"
     steps:
       - name: Check PR body
         env:

.gitignore

@@ -19,6 +19,8 @@ scripts/demo-record.sh
 
 # Internal docs — never commit to public repo
 internal/
+!benchmarks/internal/
+!benchmarks/internal/**
 
 # Benchmark results (large, ephemeral — commit RESULTS.md summaries, not raw data)
 benchmarks/longmemeval/results_*/

AGENT.md

@@ -7,7 +7,7 @@
 **Uteke** is a local-first semantic memory engine for AI agents. Single Rust binary, fully offline, ~30ms recall. No API key, Docker, or cloud service needed.
 
 - **Repo:** `codecoradev/uteke` (remote GitHub), local clone
-- **Version:** 0.14.3
+- **Version:** 0.18.0
 - **License:** Apache 2.0
 - **Main branches:** `develop` (default branch, all PRs go here), `main` (release mirror)
 

AGENTS.md

@@ -0,0 +1,45 @@
+# AGENTS.md — uteke
+
+## Project
+
+Uteke — local-first memory engine for AI agents.
+Rust workspace: uteke-core (engine), uteke-cli, uteke-server, uteke-mcp, docgen.
+
+## Stack
+
+- **Rust** (workspace, single version in root Cargo.toml)
+- **Storage**: SQLite (rusqlite bundled) + dual-engine vector index
+- **Vector engines**: usearch (default, C++ FFI) OR vecq (quantized, zero C++ dep) —
+  runtime-selected via `UTEKE_VECTOR_BACKEND` env or `[vector] backend` in uteke.toml (#1168)
+- **Embeddings**: EmbeddingGemma Q4 ONNX (default) / OpenAI / Ollama
+- **Package manager**: BUN for any JS/tooling (not npm). Rust uses cargo.
+- **Git author**: email MUST be ajianaz@users.noreply.github.com
+
+## Branch & release rules (STANDARD — applies to all CodeCora repos)
+
+- **Branches**: `develop` = integration, `main` = release (tags cut from main).
+- **Branch naming**: `feat/`, `fix/`, `docs/`, `chore/`, `perf/`, `security/`,
+  `refactor/`, `test/`, `build/`, `ci/`. `release/x` and bare names are rejected.
+- **main only accepts PRs from** `develop` or `chore/release-*` (Source Branch check).
+- **POST-RELEASE SYNC IS MANDATORY**: after a release tag is cut from main, merge
+  `main` back into `develop` immediately (PR if develop is protected). Skipping this
+  makes the next develop→main PR conflict — paid on 0.17.0 (2026-09-06).
+- **PR body**: `## What` / `## Why` / `## Testing` headers are REQUIRED (CI enforces).
+- **CI is green + Cora bot clean** before merge; never trust one gate alone.
+- Version: single workspace version in root Cargo.toml; internal deps (uteke-core/
+  uteke-mcp) must be bumped to the same version in the release commit.
+- `docs/api-reference.md` is GENERATED (`cargo run -p docgen`) — never hand-edit.
+
+## Conflicts at release time
+
+When develop→main conflicts, do NOT hand-resolve server-side (protected branch
+rejects direct pushes). Instead: branch from the release commit, `git merge main`,
+resolve keeping develop's side (unless main carries unique content — verify), push
+the branch as `chore/release-*`, and open the release PR from it.
+
+## Source of truth for workflow standards
+
+The canonical workflow SOP (merge gate, release flow, governance) lives in the
+uteke room `codecora-workflow-standard` (namespace `codecora`) on the shared
+store. This file is the repo-enforced subset — when rules disagree, the room
+wins, and this file gets updated in the same commit that changes the rule.

CHANGELOG.md

@@ -1,5 +1,31 @@
 # Changelog
 
+## [Unreleased]
+
+## [0.18.0] - 2026-09-13
+
+Minor release. Theme: **agent-operable memory plumbing** - ingest with explicit timestamps, full room lifecycle management, and a recall payload contract that keeps benchmark harnesses honest. Retrieval behavior is unchanged from 0.17.0 (revalidated at the published config: recall@5 non-abs 0.9457, identical per-question rankings).
+
+### Added (release highlights)
+
+- **Ingest date anchors: `uteke remember --timestamp` / `import` timestamps (#1232, #1238)** - attach an explicit timestamp at write time (CLI flag + batch import field) so time-travel recall (`at`), temporal boosts, and audit chains no longer depend on ingest order. The `LMEVAL_DATE_ANCHOR` env is benchmark-harness-only and never affects the binary.
+- **Room lifecycle management (#1202/#1203)** - `POST /room/rename|/room/update|/room/memory/move`, CLI `uteke room rename|update|move-memory`, MCP `uteke_room_rename|uteke_room_update|uteke_room_memory_move`; schema v19 adds additive `rooms.description` (backward-compatible exports); `Uteke::rename_room` rewrites registry + all room references in ONE transaction. Plus `uteke update <id>` for in-place memory edits.
+- **Recall payload conformance tests (#1233, #1239)** - contract tests pinning that recall responses carry the FULL payload (no hit-count stubs) across CLI/HTTP/MCP.
+
+### Added
+
+- **Benchmarks directory restructure** — `benchmarks/` is now the single source of truth: `benchmarks/README.md` (index, policies, reproduction), `internal/` (uteke bench, re-verified on v0.17.0), `longmemeval/` (harness + committed canonical raw artifacts under `results/` + corrected baselines), `locomo/` (planned external benchmark #2). The old `benchmarks/RESULTS.md` stub (stale pre-embedding figures) is removed — real numbers live in `internal/RESULTS.md` and `docs/benchmarks.md`.
+
+### Fixed
+
+- **Default recall `min_score` lowered 0.3 → 0.0 (#1223)** — since the fusion strategy (weighted RRF) became the recall default in 0.16.0, returned scores are rank-based (RRF contribution plus salience/recency boosts, typically ~0.0-0.2) rather than cosine similarity, so the CLI default threshold of 0.3 (a cosine-era value; HTTP/MCP already default 0.0) silently filtered out almost every result — measured on embeddinggemma-q4: all 20 paraphrase probes scored 0.169-0.186 rank-1 while a legacy-distribution 0.3 default expects ~0.5+, making default `uteke recall` return empty. Thresholds remain opt-in via `[recall] min_score`, `--min`, `--strict` (0.5), or HTTP `min_score`; docs now state the score scale per strategy.
+- **LongMemEval baseline mislabel** — RESULTS.md previously attributed the 0.854/0.885 (2026-08-13) full-500 baseline to "v0.15.0 hybrid"; those figures are the **vector-only** run (hybrid was validated at 50 questions only). Corrected, with both delta bases stated (+8.9pp full-500, +9.2pp 470-non-abstention).
+- **Full-dataset aggregates** — RESULTS.md now labels all three bases explicitly: Overall (470 non-abstention), Abstention (30), and Overall (full 500 = 0.982 recall_any@5 / 0.880 strict recall_all@5); no questions silently dropped from the aggregate rows.
+- Removed stale run logs and an outdated metrics snapshot from git tracking (values superseded by recomputation from committed raw artifacts).
+
+- **Room rename, room update, and memory room-move (#1202)** — rooms now have sanctioned lifecycle ops matching the namespace family: `Uteke::rename_room(old, new)` rewrites the registry row and every `room_memories`/`room_documents` reference in ONE transaction (FK-safe insert→repoint→delete order; no `ON UPDATE CASCADE` needed), preserving title/description/namespace; `Uteke::update_room(id, title?, description?)` edits room metadata (schema v19 adds the additive `rooms.description` column; structural export/import carry it and stay backward-compatible with 5-column exports); `Uteke::move_memory_to_room(id, from, to)` moves a memory between rooms preserving the link's `author`/`role`/`joined_at` provenance (`Ok(0)` when not a member of `from`). Surfaces: HTTP `POST /room/rename|/room/update|/room/memory/move` (404 on missing room/link), CLI `uteke room rename|update|move-memory`, MCP `uteke_room_rename|uteke_room_update|uteke_room_memory_move`.
+- **`uteke update <id>` CLI** — in-place memory edit (content/tags/importance/pinned/type), closing the CLI surface gap: HTTP `PUT /memory` and MCP `uteke_update` already existed (#1202 item 5).
+
 ## [0.17.0] — 2026-09-06
 
 Minor release. Theme: **inspectable, trustworthy memory** — explain recall on
@@ -8,7 +34,7 @@ graphs, pagination metadata, and a dual-engine vector layer.
 
 ### Added
 
-- **Contradiction benchmark segment (#1172, phase 3)** — `benchmarks/longmemeval/contradiction_segment.py`: 40-topic active-store segment measuring conflict-resolution quality end-to-end. Baseline (both facts active) vs resolved (superseded): fusion winner@1 0.975 → 1.000, stale@5 0.825 → 0.000. Published in `benchmarks/longmemeval/RESULTS.md`. Also adds `uteke supersede <old> <new> [--reason]` — CLI surface parity for supersession (previously MCP/HTTP only).
+- **Contradiction benchmark segment (#1172, phase 3)** — `benchmarks/longmemeval/contradiction_segment.py`: 40-topic active-store segment measuring conflict-resolution quality end-to-end. Baseline (both facts active) vs resolved (superseded): fusion winner@1 0.850 → 1.000, stale@5 1.000 → 0.000. Published in `benchmarks/longmemeval/RESULTS.md`. Also adds `uteke supersede <old> <new> [--reason]` — CLI surface parity for supersession (previously MCP/HTTP only).
 
 - **`/list` pagination metadata (#1188)** — `POST /list` accepts `"include_meta": true` to respond with an envelope `{memories, total, has_more, next_offset}` (`next_offset` is `null` on the last page) so clients no longer blind-paginate with 100-row guesses. The default response is unchanged (bare array) — existing clients are untouched; `include_meta` is ignored in `at` (point-in-time) mode, which stays a bare array.
 

Cargo.toml

@@ -9,7 +9,7 @@ members = [
 ]
 
 [workspace.package]
-version = "0.17.0"
+version = "0.18.0"
 edition = "2024"
 license = "Apache-2.0"
 repository = "https://github.com/codecoradev/uteke"

INSTALL.md

@@ -73,6 +73,10 @@ Expand-Archive -Path uteke.zip -DestinationPath uteke
 # Move to a directory in PATH (e.g. C:\Users\you\AppData\Local\bin)
 mkdir -p C:\Users\you\AppData\Local\bin
 move uteke\uteke.exe C:\Users\you\AppData\Local\bin\
+move uteke\uteke-serve.exe C:\Users\you\AppData\Local\bin\
+move uteke\uteke-mcp.exe C:\Users\you\AppData\Local\bin\
+# Required for embeddings — release zip ships ONNX Runtime DLLs alongside the exes
+move uteke\onnxruntime*.dll C:\Users\you\AppData\Local\bin\
 
 # Add to PATH (current session)
 $env:PATH += ";C:\Users\you\AppData\Local\bin"
@@ -101,7 +105,7 @@ copy target\release\uteke.exe C:\Users\you\AppData\Local\bin\
 
 ## First Run
 
-On first `remember` command, Uteke automatically downloads the embedding model (~188MB):
+On first `remember` command, Uteke automatically downloads the embedding model (~200MB):
 
 ```bash
 uteke remember "My first memory" --tags test
@@ -165,6 +169,9 @@ Check internet connection. Model downloaded from HuggingFace:
 https://huggingface.co/onnx-community/embeddinggemma-300m-ONNX
 ```
 
+### `ONNX Runtime library not found` on `uteke remember` (notably Arch/CachyOS fresh installs)
+The release tarball/zip ships `libonnxruntime.so*` / `onnxruntime.dll` next to the binaries — keep them in the same directory as `uteke` (`~/.local/bin/` on Linux/macOS). The quick-install script does this automatically; manual tarball installs must copy the `.so`/`.dylib`/`.dll` too. Fallbacks searched: `/usr/local/lib /usr/lib /lib`, pip `onnxruntime/capi/`, `~/.cache/ort.pyke.io/`, or set `ORT_LIB_PATH=/path/to/libonnxruntime.so`. Then `uteke doctor` should report `Embedding model: embeddinggemma-q4`.
+
 Manual download:
 ```bash
 # macOS/Linux