chore(deps): bump thiserror from 1.0.69 to 2.0.18#155
Closed
dependabot[bot] wants to merge 1 commit into
Closed
Conversation
Bumps [thiserror](https://github.com/dtolnay/thiserror) from 1.0.69 to 2.0.18. - [Release notes](https://github.com/dtolnay/thiserror/releases) - [Commits](dtolnay/thiserror@1.0.69...2.0.18) --- updated-dependencies: - dependency-name: thiserror dependency-version: 2.0.18 dependency-type: direct:production update-type: version-update:semver-major ... Signed-off-by: dependabot[bot] <support@github.com>
dependabot
Bot
added
dependencies
🔍 Cora AI Code ReviewReview powered by cora-cli · BYOK · MIT |
Collaborator
|
Superseded by consolidated dependency update PR |
Contributor
Author
|
OK, I won't notify you again about this release, but will get in touch when a new version is available. If you'd rather skip all updates until the next major or minor version, let me know by commenting If you change your mind, just re-open this PR and I'll resolve any conflicts on it. |
ajianaz
added a commit
that referenced
this pull request
Jun 4, 2026
- Add install.sh at root with SHA256 checksum verification - Add sync-main job to release.yml (develop → main on every v* tag) - Add SHA256 checksums generation to release job - Enable Cora Review on PRs to main branch - Update INSTALL.md to reference main/install.sh - Update release notes template to use main/install.sh Clones cora-cli #154/#155/#156 release branch pattern. Co-authored-by: CTO Hermes <cto-hermes@users.noreply.github.com>
ajianaz
added a commit
that referenced
this pull request
Jun 4, 2026
* feat: add uteke-status pi extension Pi extension that shows uteke memory stats in the footer status bar: - 🧠 uteke: 🔥3 hot | 🟡0 warm | ❄️0 cold (65 total) - Auto-refreshes after memory operations (remember, forget, import, cleanup) - Registers /uteke-stats command to manually refresh - Registers /uteke command for quick stats/aging/doctor/tags Install: copy to ~/.pi/agent/extensions/uteke-status/ or use pi install * feat(uteke-status): v2 — cross-namespace stats + system prompt injection - Query SQLite directly for cross-namespace totals (was only showing default namespace before) - Add proper spacing between icons and numbers - Inject system prompt via before_agent_start to remind agent to use uteke actively (remember, recall, search, save summaries) - Simplified: removed /uteke command, kept /uteke-stats - Uses sqlite3 CLI (pre-installed on macOS/Linux) * feat: bulk memory operations (#43) - forget <id> — now prompts for confirmation [y/N] - forget --tag <tag> — delete all memories with a tag - forget --cold — delete all cold (>30d or never accessed) memories - forget --all — delete ALL memories in namespace - All bulk ops require --confirm flag to execute - Added BulkDeleteResult type with deleted count + IDs - Added store methods: bulk_delete_by_tag, bulk_delete_cold, bulk_delete_all, count_by_tag - Added Uteke methods: bulk_forget_by_tag, bulk_forget_cold, bulk_forget_all, store() * ci: add cora AI code review composite action (#86) - Add .github/actions/cora-review composite action (downloads cora-cli from GitHub Releases, fetches LLM secrets via Infisical OIDC, runs review on PR diff, uploads SARIF to Code Scanning, posts PR comment) - Add cora-review job to CI workflow (runs after check/fmt/clippy/test pass, only on pull_request events) - Add required permissions (security-events, pull-requests, id-token) Co-authored-by: CTO Hermes <cto-hermes@users.noreply.github.com> * fix(ci): comment default, SARIF upload optional (#87) - Post PR comment: always runs (default behavior) - SARIF upload: opt-in via upload-sarif: 'true' input - SARIF upload failure: continue-on-error (never blocks) - Blocking: only error-level findings block merge Co-authored-by: CTO Hermes <cto-hermes@users.noreply.github.com> * fix(ci): resolve latest version via API before download (#89) GitHub release download URL doesn't support 'latest' tag. Resolve to actual tag via GitHub API first. Co-authored-by: CTO Hermes <cto-hermes@users.noreply.github.com> * feat: smooth namespace switching and defaults (#82) * feat: add uteke-status pi extension Pi extension that shows uteke memory stats in the footer status bar: - 🧠 uteke: 🔥3 hot | 🟡0 warm | ❄️0 cold (65 total) - Auto-refreshes after memory operations (remember, forget, import, cleanup) - Registers /uteke-stats command to manually refresh - Registers /uteke command for quick stats/aging/doctor/tags Install: copy to ~/.pi/agent/extensions/uteke-status/ or use pi install * feat(uteke-status): v2 — cross-namespace stats + system prompt injection - Query SQLite directly for cross-namespace totals (was only showing default namespace before) - Add proper spacing between icons and numbers - Inject system prompt via before_agent_start to remind agent to use uteke actively (remember, recall, search, save summaries) - Simplified: removed /uteke command, kept /uteke-stats - Uses sqlite3 CLI (pre-installed on macOS/Linux) * feat: bulk memory operations (#43) - forget <id> — now prompts for confirmation [y/N] - forget --tag <tag> — delete all memories with a tag - forget --cold — delete all cold (>30d or never accessed) memories - forget --all — delete ALL memories in namespace - All bulk ops require --confirm flag to execute - Added BulkDeleteResult type with deleted count + IDs - Added store methods: bulk_delete_by_tag, bulk_delete_cold, bulk_delete_all, count_by_tag - Added Uteke methods: bulk_forget_by_tag, bulk_forget_cold, bulk_forget_all, store() * feat: smooth namespace switching and defaults (#45) - Add `uteke namespace list` — list all namespaces with memory counts - Add `uteke namespace stats <name>` — show stats for a namespace - Add `uteke namespace switch <name>` — set default namespace in config - Implement layered namespace resolution: CLI flag > env > config > default - Fix broken `resolve_namespace` that referenced undefined variable - Support UTEKE_NAMESPACE env var for per-session defaults - Config file default_namespace persists across sessions * feat: daemon/server mode for warm recall <50ms (#83) * feat: add uteke-status pi extension Pi extension that shows uteke memory stats in the footer status bar: - 🧠 uteke: 🔥3 hot | 🟡0 warm | ❄️0 cold (65 total) - Auto-refreshes after memory operations (remember, forget, import, cleanup) - Registers /uteke-stats command to manually refresh - Registers /uteke command for quick stats/aging/doctor/tags Install: copy to ~/.pi/agent/extensions/uteke-status/ or use pi install * feat(uteke-status): v2 — cross-namespace stats + system prompt injection - Query SQLite directly for cross-namespace totals (was only showing default namespace before) - Add proper spacing between icons and numbers - Inject system prompt via before_agent_start to remind agent to use uteke actively (remember, recall, search, save summaries) - Simplified: removed /uteke command, kept /uteke-stats - Uses sqlite3 CLI (pre-installed on macOS/Linux) * feat: bulk memory operations (#43) - forget <id> — now prompts for confirmation [y/N] - forget --tag <tag> — delete all memories with a tag - forget --cold — delete all cold (>30d or never accessed) memories - forget --all — delete ALL memories in namespace - All bulk ops require --confirm flag to execute - Added BulkDeleteResult type with deleted count + IDs - Added store methods: bulk_delete_by_tag, bulk_delete_cold, bulk_delete_all, count_by_tag - Added Uteke methods: bulk_forget_by_tag, bulk_forget_cold, bulk_forget_all, store() * feat: smooth namespace switching and defaults (#45) - Add `uteke namespace list` — list all namespaces with memory counts - Add `uteke namespace stats <name>` — show stats for a namespace - Add `uteke namespace switch <name>` — set default namespace in config - Implement layered namespace resolution: CLI flag > env > config > default - Fix broken `resolve_namespace` that referenced undefined variable - Support UTEKE_NAMESPACE env var for per-session defaults - Config file default_namespace persists across sessions * feat: daemon/server mode for warm recall <50ms (#54) - Add uteke-server crate with HTTP API (tiny_http) - Endpoints: /health, /remember, /recall, /search, /list, /forget, /stats, /namespaces - Namespace support on all endpoints via JSON body - CORS headers for cross-origin access - Graceful shutdown on SIGINT (saves index, closes DB) - Config: [server] section in uteke.toml (host, port, enabled) - Benchmark: recall ~41ms (vs 2600ms CLI cold start) - Binary: uteke-serve --host 127.0.0.1 --port 8767 * feat: auto-forget, temporal facts & contradiction detection (#51) (#84) - Add temporal metadata: deprecated, valid_from, valid_until, memory_type - Memory types: fact, procedure, preference, decision, context - Contradiction detection on remember (--detect-contradiction flag) - Auto-deprecates old conflicting memories via vector similarity - Threshold calibrated for embeddinggemma-q4 model - Prune command: uteke prune --ttl 30d --dry-run - Removes deprecated memories older than TTL - Dry-run mode for safe preview - remember --type flag for memory classification - DB migration: auto-adds new columns to existing stores - All 22 tests pass * feat: knowledge graph consolidation & deduplication (#52) (#85) - Add consolidate command: uteke consolidate --threshold 0.90 --dry-run - Detect near-duplicate memories via cosine similarity - Merge duplicates: keeps newer, removes older, updates vector index - Dry-run mode shows duplicate pairs without modifying store - Add SimilarPair and ConsolidationResult types - cosine_similarity helper for O(n²) pairwise comparison - Threshold configurable (default 0.90, recommend 0.60-0.70 for small models) - All 22 tests pass * feat: CLI auto-routes to server when enabled (#90) - CLI detects running uteke-serve via HTTP health check - When server.available, routes remember/recall/search/list/stats/forget via HTTP - Recall latency: 21ms (vs 946ms CLI cold start) — 45x faster - Fallback: if server not running, CLI works as before (cold start) - Config: [server] enabled = true in uteke.toml - All 22 tests pass * chore: prepare v0.0.4 release (#91) * chore: prepare v0.0.4 release - Bump version 0.0.3 → 0.0.4 - CHANGELOG: add v0.0.4 with all new features - README: add server mode, consolidate, prune, namespace switch, fix 256d embedding - CLI reference docs: add forget bulk, consolidate, prune, namespace, server mode - Configuration docs: add server section, namespace resolution - Roadmap: add v0.0.4 completed items, update next/future - Multi-agent docs: add namespace switching section - Release workflow: build both uteke + uteke-serve, updated release notes * chore: fix clippy warnings, add cora review config - Fix all clippy warnings across workspace (uteke-cli, uteke-core, uteke-server) - Fix rustfmt issues across workspace - Remove unused Arc import in uteke-server - Fix cosine_similarity placement after test module in uteke-core - Fix config loop indexing in uteke-cli - Add .cora.yaml with custom provider (glm-5.1) - Install cora pre-commit hook (warn mode) - Fix strip fallback in release workflow - Add config migration docs to website * fix(ci): cora review exit code handling (#92) - Update cora-cli to v0.1.2 (fixes spurious exit code 2) - Add || true to cora review step so SARIF is always generated - Blocking check still works via SARIF level analysis * fix(ci): proper cora review error handling (#93) - Replace || true with explicit exit code handling - Exit code 0 (clean) and 2 (findings found) both proceed - Other exit codes show ::warning with stderr log - Empty SARIF produces warning instead of silent failure - No more 2>/dev/null — errors are visible when unexpected * fix(ci): dynamic release notes from CHANGELOG.md (#94) - Extract changelog section via awk instead of hardcoded body - Filter out bottom comparison links - Printf-based footer to avoid YAML indentation leaking - Warning annotation if CHANGELOG section not found - No more manual workflow edits per release Co-authored-by: CTO Hermes <cto-hermes@users.noreply.github.com> * feat(core): add UTEKE_HOME env var for custom data directory (#106) - Add public uteke_home() in uteke-core — resolves data directory - UTEKE_HOME set: uses that path directly (e.g. /data) - UTEKE_HOME unset: falls back to ~/.uteke (zero breaking change) - engine.rs: model path uses uteke_home() instead of dirs::home_dir() - lib.rs: doctor model check uses uteke_home() - uteke-server: DB path uses uteke_core::uteke_home() - Remove unused dirs dependency from uteke-server Closes #95 Co-authored-by: CTO Hermes <cto-hermes@users.noreply.github.com> * chore(docs): update roadmap with v0.0.5-v0.0.7 milestones (#103) - v0.0.5: Docker & Deployment (5 issues) - v0.0.6: Network & Auth (4 issues) - v0.0.7: Cloud MVP (future) - Backlog reorganized (6 items) - Color-coded sections per version - Descriptions for each milestone Co-authored-by: CTO Hermes <cto-hermes@users.noreply.github.com> * feat(server): read uteke.toml config + smart server fallback (#107) * feat(server): read uteke.toml config, default host 0.0.0.0 Server now reads [server] section from uteke.toml: - Layered resolution: uteke_home/uteke.toml → cwd/.uteke/uteke.toml - Default host changed: 127.0.0.1 → 0.0.0.0 (Docker-compatible) - CLI --host/--port override config values - Added toml dependency for config parsing Also fixes smart server fallback (#104): - CLI commands not supported via HTTP auto-fallback to local store - No more error 'This command requires local store' - Log message: 'Command not supported via server, using local store' Closes #96 Closes #104 * fix(server): resolve type mismatch in load_uteke_toml path array * style(server): apply cargo fmt --------- Co-authored-by: CTO Hermes <cto-hermes@users.noreply.github.com> * feat(server): API parity — expanded remember, GET /memory endpoint (#108) Expanded POST /remember to accept optional fields: - type: fact, procedure, preference, decision, context - valid_from, valid_until: temporal facts - detect_contradiction: auto-deprecate conflicting memories Added GET /memory?id=<uuid> endpoint: - Returns full memory by ID - 404 if not found Updated --help to document new endpoint. Closes #105 Co-authored-by: CTO Hermes <cto-hermes@users.noreply.github.com> * feat(docker): add multi-stage Dockerfile and .dockerignore (#109) Dockerfile: - Stage 1: download release binary + embedding model from HuggingFace - Stage 2: runtime only (debian:bookworm-slim + binaries + model) - ENV UTEKE_HOME=/data for volume mount - EXPOSE 8767, ENTRYPOINT uteke-serve - ARG VERSION and TARGET for build flexibility .dockerignore: - Excludes .git, target, node_modules, website, docs Closes #97 Co-authored-by: CTO Hermes <cto-hermes@users.noreply.github.com> * feat(ci): Docker image build & push on release (#110) * feat(docker): add multi-stage Dockerfile and .dockerignore Dockerfile: - Stage 1: download release binary + embedding model from HuggingFace - Stage 2: runtime only (debian:bookworm-slim + binaries + model) - ENV UTEKE_HOME=/data for volume mount - EXPOSE 8767, ENTRYPOINT uteke-serve - ARG VERSION and TARGET for build flexibility .dockerignore: - Excludes .git, target, node_modules, website, docs Closes #97 * feat(ci): add Docker image build & push to release workflow Multi-arch Docker build (linux/amd64 + linux/arm64): - Triggered after release job completes - Builds from Dockerfile with VERSION and TARGET build args - Pushes to ghcr.io/ajianaz/uteke with semver tags + latest - Uses GITHUB_TOKEN for GHCR auth (no new secrets needed) - QEMU + buildx for cross-platform Closes #99 --------- Co-authored-by: CTO Hermes <cto-hermes@users.noreply.github.com> * chore: prepare v0.0.5 release (#111) Co-authored-by: CTO Hermes <cto-hermes@users.noreply.github.com> * fix(docker): remove pinned digest and update default version (#112) Co-authored-by: CTO Hermes <cto-hermes@users.noreply.github.com> * fix(docker): remove pinned digest from runtime stage too (#113) Co-authored-by: CTO Hermes <cto-hermes@users.noreply.github.com> * fix(docker): add curl retry and timeout for CI reliability (#114) Co-authored-by: CTO Hermes <cto-hermes@users.noreply.github.com> * fix(ci): move downloads to host runner — buildx cannot access network (#115) Buildx docker-container driver cannot reach github.com/huggingface.co, causing consistent curl exit code 1. Move all downloads to CI host steps and pass files via Docker context COPY instead. - Dockerfile: COPY binaries/ and models/ from context, select by TARGETARCH - release.yml: download step before build, single multi-platform push Co-authored-by: CTO Hermes <cto-hermes@users.noreply.github.com> * fix(ci): use build artifacts + lazy runtime model download (#116) Primary: Docker job now depends on build-release (not release), gets binaries via actions/download-artifact@v4 — no curl to github.com. Fallback: Entrypoint checks if model exists at startup, downloads if missing — works even if CI model bake step fails. Co-authored-by: CTO Hermes <cto-hermes@users.noreply.github.com> * fix(ci): find artifacts recursively — merge-multiple adds subdir (#117) Co-authored-by: CTO Hermes <cto-hermes@users.noreply.github.com> * fix(ci): remove --strip-components from tar extraction (#118) The tar.gz archives contain uteke + uteke-serve at root level (no directory prefix). --strip-components=1 was stripping the filenames themselves, resulting in empty extraction. Co-authored-by: CTO Hermes <cto-hermes@users.noreply.github.com> * refactor: adopt cora-review v2 — standalone workflow + hardened action (#119) - Extract cora-review into its own workflow (.github/workflows/cora-review.yml) with dedicated concurrency group, 10-minute timeout, and scoped permissions - Harden action.yml: pinned action SHAs, checksum verification on cora-cli download, retry logic for version resolution, empty-result handling - Remove cora-review job and elevated permissions from ci.yml (CI pipeline now contains only Rust check/fmt/clippy/test/build) Co-authored-by: CTO Hermes <cto-hermes@users.noreply.github.com> * fix: v0.0.6 patch & hardening (#146) * fix: v0.0.6 patch & hardening — CI, Docker, API correctness Closes #122, #123 (already done on develop), #124, #125, #126, #145 - ci.yml, release.yml: add FORCE_JAVASCRIPT_ACTIONS_TO_NODE24 (#125) - ci.yml, release.yml: remove unused musl-tools install (#145) - release.yml: remove musl-tools apt step (#145) - Dockerfile: add non-root USER directive with uteke user (#126) - Dockerfile: create /data with correct ownership before USER switch - types.rs: add #[serde(skip_serializing)] on Memory.embedding (#124) - lib.rs: persist vector index after import() completes (#124) - vector.rs: improve expect() panic message with dims and root cause hint (#124) - .github/dependabot.yml: add Dependabot for cargo, github-actions, docker (#122) * fix(test): update test_memory_types_serialization for skip_serializing embedding The embedding field now uses #[serde(skip_serializing)], so the test verifies that embedding is excluded from JSON output and that deserialization produces an empty embedding (expected — embeddings are populated programmatically via ONNX, not from JSON). * fix: add serde(default) to skip_serializing embedding skip_serializing without default causes deserialization error when embedding field is missing from JSON. Adding default allows graceful fallback to empty vec. * release: v0.0.6 — patch & hardening - CHANGELOG: add v0.0.6 entry, reorder all versions chronologically, consolidate link refs - Cargo.toml: bump 0.0.5 → 0.0.6 --------- Co-authored-by: CTO Hermes <cto-hermes@users.noreply.github.com> * feat: v0.0.7 — core stability (#157) * feat: v0.0.7 — core stability (tag refactor, config wiring, tests) #120 Tag storage: refactor all LIKE-based tag queries to json_each() - list(), unique_tags(), tags_with_counts(), count_tag(), rename_tag(), delete_tag(), bulk_delete_by_tag(), count_by_tag - tags_with_counts: N+1 → single GROUP BY query - unique_tags: in-Rust JSON parse → SQL json_each() - 11 new tag-related tests #127 Config wiring: pass tier thresholds from config to core - Add TierConfig struct (hot_days, warm_days, hot_boost) - Uteke::open_with_tier() accepts custom tier config - MemoryTier::from_last_accessed() now takes configurable thresholds - main.rs passes config.tier.* to core - 7 new config tests (merge_from_file_all_sections, expand_tilde, etc.) #129 Test coverage: +44 tests (22 → 66 in store, 8 → 25 in lib, 4 → 15 in config) - Bulk operations: bulk_delete_by_tag, bulk_delete_cold, bulk_delete_all - Store operations: rename_tag, delete_tag, deprecate, tier_counts - Edge cases: search_content, find_aged, find_similar, pagination - Double insert, namespace isolation, empty results #140 Namespace CLI: already implemented (NamespaceCommands::List) Closes #120, #127, #129, #140 * fix: update old tests for new MemoryTier::from_last_accessed() signature - Pass hot_days=7, warm_days=30 to old tier tests - Remove dead count_tag() method (replaced by count_by_tag) * fix: update all callers for tier_counts/bulk_delete_cold new signatures Old tests called tier_counts(None) and bulk_delete_cold(None) with 1 arg, but sub-agent changed signatures to accept hot_days/warm_days. Updated 5 test call sites. * fix: correct test_path_in_memory and test_search_content_edge_cases - test_path_in_memory: rusqlite may return Some for :memory: DB path - test_search_content_edge_cases: SQLite LIKE is case-insensitive by default * docs: fix configuration section in README (#128) - Fix config search paths: .uteke/uteke.toml + ~/.uteke/uteke.toml - Fix config format: proper TOML sections matching actual Config struct - Remove non-existent --config flag from CLI reference - Add all config sections: store, embedding, tier, logging, server --------- Co-authored-by: CTO Hermes <cto-hermes@users.noreply.github.com> * release: v0.0.7 — core stability (#158) CHANGELOG: add v0.0.7 entry with all changes Cargo.toml: bump 0.0.6 → 0.0.7 Co-authored-by: CTO Hermes <cto-hermes@users.noreply.github.com> * chore(deps): consolidate Dependabot bumps — 10 deps in 1 PR (#159) Cargo deps: - thiserror 1 → 2 (API unchanged for our usage) - dirs 5 → 6 (home_dir() unchanged) - toml 0.8 → 1 (from_str() unchanged) - uuid 1.23.1 → 1.23.2 (patch) GitHub Actions: - docker/login-action v3 → v4 - docker/metadata-action v5 → v6 - actions/upload-artifact v4 → v7 - Infisical/secrets-action v1.0.9 → v1.0.16 - actions/setup-node v4 → v6 - cloudflare/wrangler-action v3 → v4 All changes reviewed — no breaking API impact for our usage. Co-authored-by: CTO Hermes <cto-hermes@users.noreply.github.com> * feat: release branch setup — main as release-only mirror (#168) - Add install.sh at root with SHA256 checksum verification - Add sync-main job to release.yml (develop → main on every v* tag) - Add SHA256 checksums generation to release job - Enable Cora Review on PRs to main branch - Update INSTALL.md to reference main/install.sh - Update release notes template to use main/install.sh Clones cora-cli #154/#155/#156 release branch pattern. Co-authored-by: CTO Hermes <cto-hermes@users.noreply.github.com> --------- Co-authored-by: CTO Hermes <cto-hermes@users.noreply.github.com>
ajianaz
added a commit
that referenced
this pull request
Jun 17, 2026
, #147, #155) Two real issues from review: 1. base_url validation (#155): OpenAiEmbedder/OllamaEmbedder stored base_url verbatim and only failed later during the HTTP call. Add crate::embed::validate_base_url() — rejects empty, schemeless, or unparseable URLs at construction time so callers see a clear Validation error instead of a confusing 404 later. Called from both OpenAiEmbedder::new and OllamaEmbedder::new. 2. Dimensions param (#146): OpenAiEmbedder accepted a dims field but never sent it in the request body. A user who set dims=3072 with a model defaulting to 1536 would silently get 1536d vectors from the API while the index believed it was 3072d. Now we include 'dimensions: N' in the request JSON when dims > 0. text-embedding-3-* honor this field; older models ignore unknown fields so this is safe to send unconditionally. Docs: - configuration.md Azure OpenAI example had an unclosed quote (#156). Fixed. Tests: +1 (rejects_invalid_base_url) covering schemeless, empty, and unparseable URLs. 187 core tests pass. Clippy clean. fmt clean.
ajianaz
added a commit
that referenced
this pull request
Jun 17, 2026
* feat: OpenAI + Ollama embedding backends (#337) Three backends now supported via [embedding] backend config: - onnx (default, fully offline, EmbeddingGemma Q4 768d) - openai (text-embedding-3-small 1536d default, Azure-compatible) - ollama (nomic-embed-text 768d default, local server) New files: - crates/uteke-core/src/embed/openai.rs (OpenAiEmbedder + 5 tests) - crates/uteke-core/src/embed/ollama.rs (OllamaEmbedder + 5 tests) Core changes (uteke-core): - EmbedderEnv resolver reads backend config from env vars without pulling the CLI config crate into core. - ensure_embedder() dispatch now constructs OpenAI/Ollama backends with defaults when model/base_url/dims are empty. - Dim mismatch detection: if backend dims != index dims on first embed call, return a clear Validation error pointing at 'uteke repair' instead of silently producing garbage vectors. - VectorIndex::dims() exposed (wraps usearch Index::dimensions()). - reqwest 0.12 gains 'json' feature (always included, no feature flag). CLI config changes (uteke-cli): - EmbeddingConfig extended: api_key, base_url, dims. - SUPPORTED_BACKENDS now [onnx, openai, ollama]. - merge_from_file handles the 3 new fields. - apply_env_overrides handles UTEKE_EMBEDDING_BACKEND/MODEL/API_KEY/ BASE_URL/DIMS. API key falls back to OPENAI_API_KEY. - +6 new config tests (full openai merge, env precedence, fallback, invalid dims). Docs: - configuration.md: expanded Embedding Backend section with backend-specific defaults table, Azure OpenAI recipe, and a dim mismatch error example. Env var table extended with the 5 new UTEKE_EMBEDDING_* vars. - CHANGELOG entry. Acceptance criteria (from issue #337): [x] OpenAiEmbedder implements Embedder trait with HTTP call [x] OllamaEmbedder implements Embedder trait with HTTP call [x] Config [embedding] backend/api_key/base_url/model/dims supported [x] Env vars override config (UTEKE_EMBEDDING_*) [x] ONNX remains default (backward compatible) [x] Dim mismatch detection with clear error message [x] reqwest always included (no feature flags) [x] Documentation updated (configuration.md) [x] Tests for backend selection logic 212 tests pass. Clippy clean. cargo fmt clean. Closes #337. * fix(embedding): plumb uteke.toml settings into backend dispatch CodeCora found that lazy backend dispatch read only env vars via EmbedderEnv::resolve, silently ignoring [embedding] api_key/base_url/ model/dims from uteke.toml. Users who set those in config (without env vars) would see backend defaults used instead. Fix: - Replace EmbedderEnv with a public EmbeddingSettings struct + a resolve_with_defaults() helper that layers env vars on top of caller-supplied config values. - Add Uteke::open_with_embedding() constructor that takes the merged CLI config as EmbeddingSettings. Stored on the Uteke struct so ensure_embedder() can read it at lazy-init time. - finish_open() gains the embedding_settings param. All other constructors forward EmbeddingSettings::default() (no behavior change for library users who don't supply settings). - VectorIndex::dims() exposed for dim-mismatch detection. - CLI main.rs now calls open_with_embedding() and forwards config.embedding.{api_key,base_url,model,dims} from uteke.toml. +3 core regression tests covering: defaults-empty, env-overrides-config, config-only-when-env-absent. Ensures the plumbing precedence (env > uteke.toml > backend default) is locked in. 215 tests pass. Clippy clean. cargo fmt clean. * fix(embedding): validate base_url + send dimensions param (CodeCora #146, #147, #155) Two real issues from review: 1. base_url validation (#155): OpenAiEmbedder/OllamaEmbedder stored base_url verbatim and only failed later during the HTTP call. Add crate::embed::validate_base_url() — rejects empty, schemeless, or unparseable URLs at construction time so callers see a clear Validation error instead of a confusing 404 later. Called from both OpenAiEmbedder::new and OllamaEmbedder::new. 2. Dimensions param (#146): OpenAiEmbedder accepted a dims field but never sent it in the request body. A user who set dims=3072 with a model defaulting to 1536 would silently get 1536d vectors from the API while the index believed it was 3072d. Now we include 'dimensions: N' in the request JSON when dims > 0. text-embedding-3-* honor this field; older models ignore unknown fields so this is safe to send unconditionally. Docs: - configuration.md Azure OpenAI example had an unclosed quote (#156). Fixed. Tests: +1 (rejects_invalid_base_url) covering schemeless, empty, and unparseable URLs. 187 core tests pass. Clippy clean. fmt clean. * fix(embedding): add UTEKE_ALLOW_DIM_MISMATCH escape hatch for repair CodeCora finding #154: the dim-mismatch guard blocked any embed-based operation once a user switched backends on an existing store, including 'uteke repair' which is the recovery path. Chicken-and-egg. Fix: skip the check when UTEKE_ALLOW_DIM_MISMATCH=1 is set. The error message now points at this flag directly: 'Rebuild the index (UTEKE_ALLOW_DIM_MISMATCH=1 uteke repair) or switch backend.' Docs: configuration.md dim-mismatch section shows the exact recovery incantation. 187 core tests pass. Clippy clean. fmt clean. * fix(embedding): treat empty env vars as unset + Azure api-version doc CodeCora finding: std::env::var returns Ok("") for explicitly empty variables, so setting UTEKE_EMBEDDING_API_KEY="" would silently overwrite a non-empty api_key from [embedding] in uteke.toml. Fix: in both lib.rs::EmbeddingSettings::resolve_with_defaults and cli/config.rs::apply_env_overrides, empty env-var values are now filtered out via .filter(|v| !v.is_empty()) before falling through to the config-provided value. Docs: the Azure OpenAI example now includes the required api-version query param (?api-version=2024-10-21) since Azure rejects requests without it. +1 core regression test (empty env does not overwrite config). 188 core tests pass. Clippy clean. fmt clean.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Bumps thiserror from 1.0.69 to 2.0.18.
Release notes
Sourced from thiserror's releases.
... (truncated)
Commits
dc0f6a2Release 2.0.180275292Touch up PR 4433c33bc6Merge pull request #443 from LucaCappelletti94/master995939cReproduce issue 44221653d1Made clippy lifetime allows conditional45e5388Update actions/upload-artifact@v5 -> v6386aac1Update actions/upload-artifact@v4 -> v5ec50561Update actions/checkout@v5 -> v6247eab5Update name of empty_enum clippy lint91b181fRaise required compiler to Rust 1.68Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting
@dependabot rebase.Dependabot commands and options
You can trigger Dependabot actions by commenting on this PR:
@dependabot rebasewill rebase this PR@dependabot recreatewill recreate this PR, overwriting any edits that have been made to it@dependabot show <dependency name> ignore conditionswill show all of the ignore conditions of the specified dependency@dependabot ignore this major versionwill close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)@dependabot ignore this minor versionwill close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)@dependabot ignore this dependencywill close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)