Merged
Conversation
Bot proxy endpoints (/chat, /chat/stream) previously required auth unconditionally, while other endpoints like /find use get_request_context which allows dev mode without API keys. This caused inconsistent behavior where ov find worked but ov chat failed with 401. Now bot endpoints use the same get_request_context dependency for consistent auth behavior across all endpoints. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Update URI utilities for better path handling - Improve memory type registry and updater - Enhance session extraction context provider - Update memory prompt templates Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- bot/ov_server.py: pass message timestamp to session.add_message() - client/session.py: pass created_at to underlying client - sync_client.py: add created_at parameter to add_message() When created_at is not provided, the server will use current time. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…ground tasks - Add end_time check in tracer.info/set/error to avoid "Tried calling _add_event on an ended span" - Add AsyncioInstrumentor to auto-create child spans for asyncio.create_task - Add @tracer decorator to _run_memory_extraction for background task tracing - Add opentelemetry-instrumentation-asyncio dependency Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
|
Failed to generate code suggestions for PR |
The openviking/telemetry/langfuse.py was not being imported anywhere in the openviking package. Project uses bot/vikingbot/integrations/langfuse.py instead. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Remove unused sys import - Sort import block alphabetically - Remove unused span variable Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- 合并两个方法，统一模板渲染逻辑 - 添加extract_context参数支持模板渲染 - 添加[content_template]日志标签用于调试events渲染问题 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
1. tools.py: 导入缺失的 truncate_content 函数和 ToolContext 类型 2. tools.py: 修复 f-string 无占位符的问题 3. test_memory_utils.py: 删除引用不存在函数的测试 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This reverts commit ede6192.
- 添加 soul.yaml 和 identity.yaml 记忆模板 - 在 MemoryField 中添加 init_value 字段支持默认值 - 在 directories.py 中添加初始化逻辑，当字段配置了 init_value 时自动创建记忆文件 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- 从 JSON 数据 conversation 提取对话最晚时间作为问题时间
- prompt 注入 "Current date: {time}" 提升时间相关问题回答质量
- 添加 --question-index 参数支持单题测试
- 添加 --skip-import 参数支持跳过导入步骤
- 新增 import_and_eval_one.sh 脚本用于单题测试
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- 简化 StructuredMemoryOperations 相关代码，移除旧版 fallback - 修复 Jinja2 渲染 StrPatch 对象时调用 __str__ 的问题 - 统一文件存在/不存在逻辑，merge_op.apply 自动处理 current_value=None - serialize_with_metadata 统一处理模板渲染 - 移除 _extract_replace_from_patch 等冗余方法 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- 优化 memory_updater.py 文件写入逻辑 - 修复模板渲染问题 - 更新 memory 模板文件 (entities, events, preferences, profile, soul, tools) - 修复 language.py 相关问题 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- 使用 question_id（如 conv-26_qa0）作为 session_id，每个 query 完全独立 - 恢复 --threads 参数，默认值 20，控制并发度 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- import_and_eval_one.sh: 添加 --skip-import 参数支持跳过导入 - run_eval.py: is_invalid 改为按问题内容匹配，支持 errors.json - 调整 is_invalid 字段位置到 result 后面 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- 默认路径从 test_data 改为 data - 添加 import_and_eval_one.sh 使用说明 - 添加 --skip-import 开关说明 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
qin-ctx
approved these changes
Apr 6, 2026
zeattacker
pushed a commit
to zeattacker/OpenViking
that referenced
this pull request
Apr 10, 2026
Upstream catch-up from volcengine/OpenViking main at a18d4b9. Auto-merged (24 files): README, bot/config, crates/ov_cli, session/session.py, session/compressor{,_v2}.py, session/memory_deduplicator.py, memory_extractor.py, schema_model_generator.py, utils/uri.py, storage/viking_fs.py, storage/transaction/lock_manager.py, storage/queuefs/semantic_processor.py, retrieve/hierarchical_retriever.py, server/routers/content.py, server/routers/search.py, core/directories.py, prompts/templates/memory/preferences.yaml, examples/openclaw-plugin/{index,text-utils}.ts, examples/ov.conf.example, openviking_cli/utils/config/{memory_config,open_viking_config}.py, docs/en/concepts/08-session.md. Hand-resolved conflicts (14 files): Memory subsystem: - entities.yaml: adopt upstream category field + event linking, keep our brain-hardening rules (canonical names, Aliases section, one-card-per-entity). Removed upstream's Caroline LoCoMo leak. - events.yaml: adopt upstream year/month/day folder template. - memory_updater.py: keep dev's two-function _apply_write/_apply_edit structure with collision-safe entity writes, port get_year/get_month/get_day helpers to ExtractContext for events.yaml template to work. - session_extract_context_provider.py: keep dev's [Keywords] bug fix (_derive_search_keywords) — upstream volcengine#1159 did not address this. - extract_loop.py: keep dev's small-model extraction path. Dev's hard_cap iteration extension is more sophisticated than upstream's version. Storage: - collection_schemas.py: combine dev's BGE-M3 truncation (30k chars) with upstream volcengine#1301 embed_compat async wrapper. - queuefs/semantic_dag.py: combine dev's old_summary hash comparison with upstream's defensive null check on cached summary. - queuefs/semantic_queue.py: keep dev's 300-second dedup window with _TrackedSemanticRequest dataclass. - utils/summarizer.py: take upstream — convergent fix, upstream is superset of our context_type centralization plus resource cover handling. Models: - openai_vlm.py: merge timeout signature (float | None = 60.0). Plugin (TypeScript): - config.ts: keep dev's profileInjection/recallFormat/alignment fields, add upstream's bypassSessionPatterns deprecation alias. - client.ts: merge addSessionMessage signature (sessionId, role, content, parts?, agentId?, createdAt?), keep dev's createSession(), combine body building with optional created_at. - context-engine.ts: keep dev's driftDetector + alignment, add upstream's isBypassedSession helper + bypass early-return in doCommit/afterTurn. Drop upstream's inline afterTurn commit block — dev routes through doCommitOVSession. Update addSessionMessage call to 6-arg form. - memory-ranking.ts: keep dev's multi-slot recall + tool-experience separation architecture. Deferred upstream changes: - volcengine#1221 agfs→ragfs Rust rewrite: new Rust crates land but Python code unchanged; pyagfs handles auto-fallback via RAGFS_IMPL env var. - volcengine#1159 memory_updater unified operations refactor: kept dev's separate write/edit function structure to avoid cascading schema changes through extract_loop and related files. Critical security + bug fixes preserved from upstream: - volcengine#1319 leaked token removed from settings.py - volcengine#1133 SSRF hardening on HTTP resource ingestion - volcengine#1297 sanitize and cap recall queries (wraps our plugin recall) - volcengine#1277 configurable embedding circuit breaker - volcengine#1279 trusted mode without API key restricted to localhost - volcengine#1211 PID lock recycle + ownership checks + compressor refs - volcengine#1182 task API ownership leakage fix in content.py - volcengine#1301 embedder async contention reduction - volcengine#1226 prevent VLM blocking startup hang in redo recovery - volcengine#769/volcengine#792 queuefs dedupe memory semantic parent enqueues Critical dev fixes preserved: - [Keywords] placeholder root-cause fix (brain-hardening plan phase 1) - Episodic memory v2 (retrieval scope, category boosts, archive filter) - Qwen-9B small-model extraction compatibility - Plugin recall refactor (tool-experience separation, multi-tier recall) - Context-type centralized inference (convergent with upstream fix) - BGE-M3 embedding input truncation - URI normalization + collision-safe entity writes Merged on dev-local2 safety branch. Next steps: build + test before fast-forwarding dev. Co-Authored-By: claude-flow <ruv@ruv.net>
zeattacker
pushed a commit
to zeattacker/OpenViking
that referenced
this pull request
Apr 10, 2026
The initial merge of upstream 'Feature/memory opt' (volcengine#1159) auto-merged a unified operations schema into utils/uri.py and utils/content.py while our TAKE DEV choice left memory_updater.py / extract_loop.py expecting the old split schema. Runtime smoke test revealed cascading API mismatches: - 'ResolvedOperations' has no 'write_operations' (AttributeError) - serialize_with_metadata() arg order mismatch (TypeError on write) - LLM output validation would fail on missing reasoning/edit_overview_uris fields (upstream commented them out in schema_model_generator) Two parallel analysis agents (arch map of volcengine#1159 + call-graph audit) both converged on: TAKE DEV for schema-breaking changes, KEEP upstream for purely-additive improvements. This commit is the resulting minimal diff. Reverted to dev (schema-breaking bits): - utils/uri.py: ResolvedOperations keeps write_operations/edit_operations split (load-bearing for brain-hardening _MERGE_ON_COLLISION_TYPES) - utils/content.py: serialize_with_metadata(content, metadata) positional signature matches all 5 call sites (memory_updater x2, memory_type_registry x1, content_write x2 — upstream's new content_write also uses this form) - schema_model_generator.py: uncommented 'reasoning' and 'edit_overview_uris' Pydantic field definitions — extract_loop.py:196 requires both Patched to bridge upstream additive + dev signature: - memory_type_registry.py::initialize_memory_files — render content_template inline with jinja2, then call dev's serialize_with_metadata(rendered, metadata) instead of upstream's serialize_with_metadata(metadata, content_template=, extract_context=) Kept upstream wins untouched (already in merged state, fully additive): - MemoryTypeRegistry.initialize_memory_files() method (volcengine#1159) - MemoryTypeRegistry.__init__(load_schemas=True) auto-loading - ExtractContext.get_year/get_month/get_day helpers (used by events.yaml) - openviking/telemetry/tracer.py module - compressor_v2 v2_lock_retry config + tracer (volcengine#1275) - merge_op/base.py get_first_replace() helper - merge_op/patch.py None-handling branch - MemoryField.init_value - soul.yaml / identity.yaml prompt templates - events.yaml year/month/day folder structure - AGFS Rust binding (volcengine#1221) — confirmed active at runtime Smoke test result (conv-26 session_1, qwopus-4b, temp=0.1): Written: 3, Edited: 0, Deleted: 0, Errors: 0 4 memories total (2 entities, 1 episode, 1 other) Historical baseline variance range: 4-10 memories, median ~6. Result within normal LLM variance. Co-Authored-By: claude-flow <ruv@ruv.net>
zeattacker
pushed a commit
to zeattacker/OpenViking
that referenced
this pull request
Apr 10, 2026
Smoke test after the surgical volcengine#1159 integration (8b43ae7) showed a systematic regression: only entities + episodes extracted (4 memories), missing events/preferences/profile across 3 deterministic runs. Root cause (from server log inspection of final_operations JSON): 1. events.yaml — upstream changed filename_template to "{{get_year(ranges)}}/{{get_month(ranges)}}/{{get_day(ranges)}}/{{event_name}}.md" The get_year/get_month/get_day helpers on ExtractContext derive year/ month/day from the first message's `created_at` attribute via _first_message_time(). The LoCoMo ingest path does not populate created_at, so all three helpers return None and the filename renders as None/None/None/foo.md — URI resolution rejects it silently. LLM emits the event correctly; the template can't render. 2. preferences.yaml — upstream changed filename_template from "{{ user }}_{{ topic }}.md" (flat, proven working in dev) to "{{ user }}/{{ topic }}.md" (subdirectory per user) Both reference the same `user` + `topic` fields. In practice the switch coincides with Qwen-4B emitting `"preferences": []` in the extraction output — possibly because the subdir form in the rendered schema prompt confuses the small model, possibly incidental. Either way dev's form is proven in production, upstream's is not. Decision: revert both filename templates to dev's proven form. Keep upstream's additive "Record content that mentions time" rule and the shorter example text in events.yaml — those don't touch rendering. Smoke test after fix (3 deterministic runs, qwopus-4b, temp=0.1): 7 memories total entities 2 episodes 1 events 1 preferences 1 profile.md 1 vs baseline (reverted_smoke 2026-04-09): 6 memories, same 5 categories. +1 over baseline, 3/3 runs identical. Co-Authored-By: claude-flow <ruv@ruv.net>
zeattacker
pushed a commit
to zeattacker/OpenViking
that referenced
this pull request
Apr 10, 2026
…mples
The post-merge entities.yaml prompt I wrote had two related bugs surfaced
during the 19-session LoCoMo bench:
GET viking://user/.../memories/entities/entities/people/melanie.md → 404
GET viking://user/.../memories/entities/entities/pets/oscar.md → 404
GET viking://user/.../memories/entities/events/2023/06/02/x.md → 404
Root cause: entity files live at entities/<category>/<name>.md (per the
filename_template '{{ category }}/{{ name }}.md' adopted from upstream
volcengine#1159), but my prompt examples used `../entities/<category>/<name>.md`
for entity-to-entity links. From inside entities/people/caroline.md,
`..` already exits to entities/, so the link resolved to
entities/entities/people/melanie.md — doubled prefix.
Same root cause for entity-to-event links: I had `../events/<year>/<month>
/<day>/<event_name>.md` which (a) only hops one level instead of two
(needs out of <category>/ AND out of entities/) and (b) used the date-
hierarchy filename format that we explicitly reverted in commit 818f083.
Fix:
- Same-category sibling: `<other_name>.md` (no `..` needed)
- Different category: `../<other_category>/<name>.md` (1 level up)
- Event reference: `../../events/<YYYY-MM-DD>_<event_name>.md`
(2 levels up, flat events filename)
Added an explicit "Entity files live at entities/<category>/<name>.md.
Compute relative paths from THAT location" rule to anchor the LLM's
mental model.
This fix only affects new extractions — existing entity files in
production memories will still have the broken links until rewritten.
Co-Authored-By: claude-flow <ruv@ruv.net>
Ferry-200
added a commit
to Ferry-200/OpenViking
that referenced
this pull request
Apr 11, 2026
* feat(encryption): add encrypt unit test for plaintext shorter than the magic header (volcengine#1217) * fix: PID lock recycle, recall threshold bypass, orphaned compressor refs (volcengine#1211) * fix: PID lock recycle, recall threshold bypass, orphaned compressor refs 1. process_lock: verify PID is actually OpenViking on Linux by checking /proc/{pid}/cmdline. Prevents false DataDirectoryLocked when PIDs are recycled to unrelated processes after crash (Fixes volcengine#1088). 2. memory-ranking: add scoreThreshold param to pickMemoriesForInjection and filter non-leaf items below threshold. Previously low-scoring memories bypassed recallScoreThreshold when supplementing leaves (Fixes volcengine#1106). 3. compressor: catch FileNotFoundError separately in _merge_into_existing, clean up orphaned vector records so they are not retried indefinitely (Fixes volcengine#1048). * style: ruff format process_lock.py --------- Co-authored-by: JasonOA888 <JasonOA888@users.noreply.github.com> * fix(queuefs): dedupe memory semantic parent enqueues within window (volcengine#769) (volcengine#792) Repeated memory writes enqueue the same parent directory for .overview/.abstract regeneration, causing redundant VLM work. Skip duplicate enqueues for the same (account, user, agent, uri) within 45s; resource/session semantic paths unchanged. Fixes volcengine#769 Made-with: Cursor * docs: add Claude Code Memory Plugin example link and Chinese docs (volcengine#1228) Add README link for Claude Code Memory Plugin example in all language variants (EN, CN, JA) and add Chinese documentation for the plugin. * fix: prevent startup hang from blocking VLM call in redo recovery (volcengine#1226) Move redo recovery to a background task so the server starts without waiting for potentially slow VLM calls. Add 60s timeout to extract_long_term_memories to prevent indefinite hangs on individual redo tasks. Clean up the redo task on stop(). Fixes volcengine#1222 Co-authored-by: Matt Van Horn <455140+mvanhorn@users.noreply.github.com> * fix(cli): handle non-UTF-8 filenames in add-resource (volcengine#1224) Replace to_string_lossy() with to_str() to preserve valid UTF-8 filenames (Chinese, Japanese, special characters) instead of silently corrupting them. Non-UTF-8 paths now return a clear Error::InvalidPath instead of garbled output. Fixes volcengine#1018 Co-authored-by: Matt Van Horn <455140+mvanhorn@users.noreply.github.com> * test(crypto): add regression tests for decrypting short plaintext files (volcengine#1223) Add tests covering the fix in PR volcengine#1163 where decrypt() raised 'Ciphertext too short' on plaintext files shorter than 4 bytes. Tests added: - test_decrypt_empty_plaintext: raw b'' returns b'' - test_decrypt_short_plaintext_less_than_4_bytes: b'X', b'AB', b'ABC' return as-is - test_decrypt_magic_prefix_without_full_header: b'OVE1' raises CorruptedCiphertextError Co-authored-by: yc111233 <yc111233@users.noreply.github.com> * Implement request-scoped wait for write APIs (volcengine#1212) fix: request wait telemetry id fix: register request wait before enqueue add log * reorg: Rewrite agfs to ragfs with rust (volcengine#1221) * reorg: rewrite agfs with rust, and named with ragfs, keep License * reorg: rewrite agfs with rust, and named with ragfs, keep License * reorg: rewrite agfs with rust, and named with ragfs, keep License * reorg: rewrite agfs with rust, and named with ragfs, keep License * reorg: rewrite agfs with rust, and named with ragfs, keep License * reorg: rewrite agfs with rust, and named with ragfs, keep License * reorg: rewrite agfs with rust, and named with ragfs, keep License * reorg: rewrite agfs with rust, and named with ragfs, keep License * fix: grep level limit * fix: grep root * fix: import error * fix: rust code optimazation * fix: CI error * fix: CI go mod cache * fix: grep level limit * fix: CI --------- Co-authored-by: openviking <openviking@example.com> * fix: update main when release, and add docker hub push (volcengine#1229) * fix security: feat(resources): harden HTTP resource ingestion against private-network SSRF (volcengine#1133) * Harden HTTP resource ingestion against private-network SSRF * chore(ci): retrigger checks * style: fix resource service import order * fix(session): add timeout to _wait_for_previous_archive_done to prevent infinite hang (volcengine#1235) When Phase 2 of a previous archive crashes after writing messages but before writing .done or .failed.json, the next archive's memory extraction enters an infinite polling loop with no exit condition. Add a 5-minute deadline. On timeout, return False (treated as failure) so the caller writes .failed.json and the session is no longer stuck. Co-authored-by: yc111233 <yc111233@users.noreply.github.com> Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com> * fix: volcengine#1238 and volcengine#1242 and volcengine#1232 (volcengine#1243) * fix(bot): respect OPENVIKING_CONFIG_FILE even when file doesn't exist Previously, bot's config path resolution would fallback to ~/.openviking/ov.conf when OPENVIKING_CONFIG_FILE was set but the file didn't exist. This was inconsistent with server's behavior, which treats a missing env-specified config as an error. In container deployments with OPENVIKING_CONFIG_FILE=/app/ov.conf, this caused bot to potentially write auto-generated config to /root/.openviking/ov.conf instead of the intended /app/ov.conf, leading to config file path mismatches. Now bot respects the environment variable unconditionally, matching server's behavior and ensuring both components use the same config path. Fixes volcengine#1242 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com> * fix(embedder): support dimension truncation for OpenAI-compatible models Previously, when users configured a dimension (e.g., 1024) for OpenAI-compatible embedding models that don't support the 'dimensions' API parameter, the system would fail with "dimensions is currently not supported" error. Additionally, when dimension was not configured, the config layer would return a hardcoded fallback of 2048, but the actual model might return a different dimension (e.g., 1024), causing dimension validation failures. This fix implements vector truncation in OpenAIDenseEmbedder: - Removes the 'dimensions' parameter from API calls (not supported by all models) - If user configures dimension=1024 and model returns 2048, truncates to 1024 - If no dimension is configured, uses model's native dimension without truncation - Applies truncation to both single and batch embedding operations This allows users to: 1. Use OpenAI-compatible models with custom dimensions via truncation 2. Control vector dimensions for storage optimization 3. Avoid dimension mismatch errors between config and actual embeddings Fixes volcengine#1238 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com> * fix: volcengine#1232 --------- Co-authored-by: openviking <openviking@example.com> Co-authored-by: Claude <noreply@anthropic.com> * Feature/memory opt (volcengine#1159) * fix(plugin): add skills to autoRecall search scope (volcengine#1225) Include viking://agent/skills in the autoRecall search alongside user memories and agent memories. Skills stored in OpenViking are now auto-injected into context when relevant to the query. Fixes volcengine#1089 Co-authored-by: Matt Van Horn <455140+mvanhorn@users.noreply.github.com> * Revert "fix(session): add timeout to _wait_for_previous_archive_done to preve…" (volcengine#1265) This reverts commit dc15daf. * channel mention (volcengine#1272) * feat(storage): volcengine vector db support sts token (volcengine#1268) * feat(storage): volcengine vector db support sts token https://www.volcengine.com/docs/7139/1302258?lang=zh * feat(storage): volcengine vector db support sts token https://www.volcengine.com/docs/7139/1302258?lang=zh * feat(server): support host all to use dual stack netstat (volcengine#1273) * feat(auth): Restrict trusted mode without API key to localhost (volcengine#1279) * Improve memory v2 lock retry controls in compressor (volcengine#1275) * fix(security): configurable embedding circuit breaker & log suppression (volcengine#1277) * update wechat group qrcode (volcengine#1282) replace the old one. * ci: optimize API test matrix from 5 to 3 channels (volcengine#1281) - Reduce OS matrix from 5 channels to 3 channels - Use ubuntu-latest, macos-latest, windows-latest instead of specific versions - Remove redundant ubuntu-arm and intel macos variants * feat: add MiniMax-M2.7 and MiniMax-M2.7-highspeed provider support (volcengine#1284) - Update provider registry to document MiniMax-M2.7 and MiniMax-M2.7-highspeed as the recommended models (replacing the outdated M2.1 example comment) - Add explicit note that MiniMax does not support system messages (they are merged into the first user message automatically) - Update README to advertise MiniMax-M2.7 and MiniMax-M2.7-highspeed as the recommended MiniMax models with configuration instructions - Add comprehensive unit tests (17 tests) covering: - Registry keyword matching for MiniMax-M2.7 and MiniMax-M2.7-highspeed - Model prefix resolution (minimax/MiniMax-M2.7) - System message merging for both LiteLLMProvider and OpenAICompatibleProvider - Edge cases: multiple system messages, no system messages, non-MiniMax models - MINIMAX_API_KEY environment variable and international API base URL Co-authored-by: octo-patch <octo-patch@github.com> * feat(eval): add openclaw eval sh (volcengine#1287) * 增加完整的一键评测脚本 * 增加完整的一键评测脚本 * add create time in add_message (volcengine#1288) * fix(lark): add lark-oapi (volcengine#1285) * docs: add prompt guides in zh and en (volcengine#1292) * benchmark: add LoCoMo evaluation scripts for mem0 (volcengine#1290) Implements a two-phase benchmark pipeline for evaluating mem0 on the LoCoMo long-term conversation dataset (10 samples, 1540 non-adversarial QA pairs). - ingest.py: imports LoCoMo conversation sessions into mem0, using sample_id as the userId namespace. All messages use "user" role with [SpeakerName]: prefix to preserve two-person dialogue structure. Temporal context is added via a [System] prefix on each session. - eval.py: sends QA questions to an OpenClaw agent backed by the openclaw-mem0 plugin. Restarts the gateway per sample to switch the active userId, verifies the correct user is loaded before running questions, then parallelizes questions within each sample using unique session keys. Parses session jsonl to collect accurate per-turn token usage. Optionally judges answers with a Volcengine ARK LLM. - delete_user.py: utility to clear mem0 memories for given user_ids. - README.md: documents prerequisites, ingest/eval parameters, output format, and per-sample run commands. * fix(docker): restore venv tooling for ragfs build (volcengine#1295) fix(docker): preserve ragfs wheel extraction script indentation fix(docker): keep heredoc cleanup inside run shell fix(docker): terminate ragfs extraction heredoc cleanly fix(ci): split docker manifest digests by registry * Include gemini optional dependency in Docker image (volcengine#1254) The Dockerfile only includes the `bot` extra when installing dependencies, which means the `google-genai` package is missing at runtime. This causes the server to crash with `TypeError: 'NoneType' object is not callable` when users configure the Gemini embedding provider. Add `--extra gemini` to the `uv sync` commands so the Docker image ships with Gemini support out of the box. Fixes volcengine#1253 * ci: add timeout and SMTP failure notification (volcengine#1293) - Add 50 minutes timeout for api_test and oc2ov_test workflows - Add SMTP email notification on workflow failure - Email will be sent to likaisong@bytedance.com Required secrets: - SMTP_USERNAME: SMTP email username - SMTP_PASSWORD: SMTP email password/app password * fix(openclaw): sanitize and cap recall queries (volcengine#1297) Recall should use cleaned user text and avoid sending oversized prompts to retrieval. Add unit coverage for sanitization and truncation behavior. * fix(ci): repair reusable build workflow yaml blocks (volcengine#1300) * feat(ast): add Lua parser support and extractor wiring (volcengine#1286) * fix(ci): remove disallowed notify-failure action (volcengine#1302) - Remove dawidd6/action-send-mail@v4 which is not in allowed list - Keep timeout-minutes: 50 configuration - Use GitHub built-in notifications instead * fix(embedder): reduce async contention in session flows (volcengine#1301) Introduce native async embedding paths across providers, switch async retrieval/session hotspots to use them, and add a standalone mixed-load benchmark plus before/after benchmark evidence for the regression. * Feat(ovpack): recursively vectorize all imported docs (volcengine#1294) * feat(ovpack): recursively enqueue directory vectorization * feat(ovpack): recursively vectorize all imported docs * refactor(ovpack): use fixed workers for direct vectorization * refactor(ovpack): revert direct vectorization to gather * chore: drop redundant default tree args * feat: 巻加测试优化功能 (volcengine#1280) - Session ID 自动管理 - 智能等待策略 - 重试机制 - 测试数据管理 * fix(eval): OpenClaw eval, import to ov use default user (volcengine#1305) * 增加完整的一键评测脚本 * 增加完整的一键评测脚本 * 完善评测脚本 * 完善评测脚本 * 完善评测脚本 * fix(memory): batch semantic processing in _process_memory_directory to prevent CPU spikes (volcengine#1245) (volcengine#1304) * Fix ci (volcengine#1307) * fix(ci): fix build ci * build: move ragfs-python packaging into setup.py * fix bug (volcengine#1317) * fix: 优化测试关键词匹配和移除 Release Approval Gate (volcengine#1313) - 添加中英文关键词匹配：删除、无、deleted、expired、no longer - 移除 workflow 中的 release-approval job * fix: remove debug print statement in bot health check endpoint (volcengine#1310) Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com> * feat: add scenario-based API tests (volcengine#1303) * feat: add scenario-based API tests - Add scenario test framework with proper categorization - Add tests for resources_retrieval, sessions, and stability_error scenarios - Add get_task() and wait_for_task() methods to API client for async operations - Add get_session_context() method for session context retrieval - Update API test workflow name from '03' to '06' - All 14 scenario tests pass with proper business logic validation * fix: skip scenarios tests when no VLM/Embedding secrets Scenarios tests require VLM for session archival summaries and Embedding for semantic search. Skip them in basic test mode. * fix(security): remove leaked token from settings.py (volcengine#1319) - Remove tests/oc2ov_test/config/settings.py containing exposed auth token - Add settings.py to .gitignore to prevent future leaks - Users should copy settings.example.py to settings.py and fill in their own tokens * fix(resources): implement trailing slash semantics for resource URIs (volcengine#1321) add support for trailing slash rules in resource URIs to control file/directory placement update CLI, API, and documentation to reflect new URI handling semantics add comprehensive tests for all URI semantics cases * Revert "fix(resources): implement trailing slash semantics for resource URIs …" (volcengine#1322) This reverts commit dec57bd. * fix litellm embedding dimension issues (volcengine#1323) Co-authored-by: openviking <openviking@example.com> * ci: optimize runner usage with conditional OS matrix and parallel limit (volcengine#1327) - Only run full OS matrix (ubuntu-24.04, macos-14, windows-latest) on main branch pushes - PR branches use ubuntu-24.04 only to reduce runner wait time - Add max-parallel: 2 to limit concurrent job execution * fix(bot):Response language, Multi user memory commit (volcengine#1329) * 修复语种问题 * 多user提交ov * ci: remove lite and full test workflows (volcengine#1331) * docs: fix docker deployment (volcengine#1332) Co-authored-by: openviking <openviking@example.com> * docs(openclaw-plugin): add health check tools guide (volcengine#1326) Add solution for health check tools report http 404 error * fix(queue): expose re-enqueue counts in queue status (volcengine#1337) Track semantic and embedding re-enqueues as first-class queue metrics so observer output, wait_processed payloads, and telemetry summaries make retry loops visible before they escalate into hard errors. * feat(s3fs): add disable_batch_delete option for OSS compatibility (volcengine#1330) (volcengine#1333) Co-authored-by: yuanqianhe <yuanqianhe@lattebank.com> * Fix/add resource cover (volcengine#1338) * fix(resources): improve handling of resource imports and naming - Add source_name to file upload requests for preserving original filenames - Handle single-directory zip files by using their root directory directly - Support viking://resources as parent directory for imports - Split summarization for resources root imports into individual child items - Add tests for new resource import behaviors * style(tests): format test files with consistent line breaks Improve readability by applying consistent line breaks in test file patches and removing trailing whitespace * afterTurn: store messages with actual roles and skip heartbeat messages (volcengine#1340) * fix: fall back to prefix filters for volcengine path scope (volcengine#1342) Co-authored-by: haosenwang1018 <haosenwang1018@users.noreply.github.com> * fix(session): auto-create missing sessions on first add (volcengine#1348) Ensure the add-message API materializes a missing session before loading it so plugins can append the first turn without an explicit create_session call. * Fix/api test issues (volcengine#1341) * fix: make api_test more robust for CI environments - Add @pytest.hookimpl(optionalhook=True) for pytest-html hooks to fix compatibility issues - test_fs_read: skip test when AGFS service is not available - test_get_overview: skip test when overview file does not exist These changes ensure tests pass gracefully on CI servers where AGFS service may not be available or files may not exist. * ci: reduce max-parallel to 1 for better resource availability Reduce max-parallel from 2 to 1 to avoid waiting for multiple runners when GitHub-hosted runners are limited. * fix: derive context_type from URI in index_resource (volcengine#1346) index_resource() always passed the default context_type="resource" to vectorize_directory_meta() and vectorize_file(), even when indexing memory directories (URIs containing /memories/). This caused all records created via the /api/v1/content/reindex endpoint to be tagged as "resource" instead of "memory", breaking stats aggregation (which filters on context_type="memory") and search scoping. Use the existing get_context_type_for_uri() helper to derive the correct context_type from the URI and pass it through to both vectorize_directory_meta() and vectorize_file(). Co-authored-by: yc111233 <yc111233@users.noreply.github.com> Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com> * reorg: remove golang depends (volcengine#1339) * docs: fix docker deployment * reorg: remove third_party/agfs * feat(s3fs): add disable_batch_delete option for OSS compatibility Port of PR volcengine#1333 from Go version to Rust: - Add disable_batch_delete config option to S3Client - When enabled, use sequential single-object deletes instead of DeleteObjects - This is for S3-compatible services like Alibaba Cloud OSS that require Content-MD5 for DeleteObjects but AWS SDK v2 does not send it by default - Add documentation and config example for OSS * fix(s3fs): pass disable_batch_delete config from Python to Rust Add disable_batch_delete to the s3_plugin_config dict in _generate_plugin_config so that the Python config can properly control the Rust S3FS plugin's behavior. * reorg: remove third_party/agfs * reorg: remove third_party/agfs * change some docs * change some docs --------- Co-authored-by: openviking <openviking@example.com> * Feat/mem opt (volcengine#1349) * fix(memory): handle string response from VLM when tools disabled - Fix AttributeError when VLM returns string instead of VLMResponse - Fix tuple creation bug with trailing comma causing double-nested tools array Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * update * chore: replace LGBTQ example with book club in entities.yaml Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(memory): disable tools on extended iteration to prevent infinite loop When max_iterations is extended due to tool calls, disable tools for the additional iteration to ensure the extract loop terminates. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(memory): handle out-of-bounds range from LLM extraction Clamp range values to valid message indices instead of skipping, to handle cases where LLM extracts incorrect ranges. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com> * fix: openai like embedding models fix, no more matryoshka error (volcengine#1350) Co-authored-by: openviking <openviking@example.com> * 增加关闭ov的配置 (volcengine#1352) * feat: 新前端项目初始化 * style: 切换到vega风格 * feat: 新增定制化的 openapi codegen * chore: 把openapi生成的ov-client移到src/gen底下 * feat: 生成 openviking-server 的 client * feat: 搭建 openviking client 的适配层 * docs: 新增 ov-client 适配层描述 * chore: 清理项目模板 * clean: 清理模板样式 * deps: 新增 tanstack query 依赖 * style: 调整 shadcn 风格 * feat: 导入基础的shadcn组件 * feat: 复刻旧的 webconsole * fix: 修复legacy代码commit不完整的问题 * docs: readme * feat: 允许不显示传递自定义client * dev: add development environment script * feat: support tags for resource management and retrieval This commit introduces a native `tags` parameter across the entire stack (Core, API, SDK, CLI) to easily tag resources and filter them during semantic search. Changes include: - **Core & Storage**: - Write `tags` to the resource root's `.meta.json` during `add_resource`. - Read tags from `.meta.json` during async semantic processing and hoist them to the VectorDB context for indexing. - Enrich directory stats/entries in `VikingFS` (`ls`, `stat`) with `tags`. - **API & Service**: Add `tags` field to resource creation and search routes. - **Python SDK**: - Add `tags` parameter to `add_resource`. - Add `tags` shortcut parameter to `find` and `search` methods, which automatically constructs the underlying `contains` metadata filters. - **Rust CLI**: Add `--tags` flag to `ov add-resource`, `ov find`, and `ov search` commands. - **Docs**: Update English and Chinese documentation for Resources, Retrieval, and Filesystem APIs to reflect the new `tags` parameters and structures. - **Tests**: Add unit tests for resource processor meta merging, VikingFS tag reading, and HTTP client tag filtering logic. * feat(search): support tags filtering and return tags in retrieval results * refactor(tags): extract build_tags_filter helper and improve validation * chore(scripts): add fast mode toggle to bootstrap menu * fix(tags): sanitize once and isolate semantic tags context * feat: 三段式布局 + 功能页拆分 - 新增顶栏(全宽) + 左侧边栏 + 右侧内容区的三段式布局 - 将 Data 大页面拆分为独立路由: FileSystem、Find、Add Memory - 将 Access 页面迁移为独立的 Settings 页面 - 侧边栏按 Data/Ops/Access 分组显示导航项 - 提取共享工具函数到 data-utils.ts - 修复 .gitignore 中 data/ 规则误匹配子目录的问题 * feat(web-studio): add i18n support and dark/light theme toggle - Add i18next with browser language detection (zh-CN / en) - Add language switcher dropdown in header bar - Add dark/light theme toggle with animated Sun/Moon icons - Wire up next-themes ThemeProvider with class-based dark mode - Replace shadcn dark theme with default Zinc (neutral gray) - Connect sidebar labels to i18n translation keys - Fix dropdown menu forced dark styling --------- Co-authored-by: baojun-zhang <zhangbaojun.1@bytedance.com> Co-authored-by: Jason <101583541+JasonOA888@users.noreply.github.com> Co-authored-by: JasonOA888 <JasonOA888@users.noreply.github.com> Co-authored-by: Protocol Zero <257158451+Protocol-zero-0@users.noreply.github.com> Co-authored-by: 灿烂甜菜 <731426007@qq.com> Co-authored-by: Matt Van Horn <mvanhorn@users.noreply.github.com> Co-authored-by: Matt Van Horn <455140+mvanhorn@users.noreply.github.com> Co-authored-by: yc111233 <109650216+yc111233@users.noreply.github.com> Co-authored-by: yc111233 <yc111233@users.noreply.github.com> Co-authored-by: Jiahui Zhou <zhoujiahui.01@bytedance.com> Co-authored-by: MaojiaSheng <shengmaojia@bytedance.com> Co-authored-by: openviking <openviking@example.com> Co-authored-by: 13ernkastel <LennonCMJ@live.com> Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com> Co-authored-by: chenjw <chenjunwen@bytedance.com> Co-authored-by: yeshion23333 <dutao.1786@bytedance.com> Co-authored-by: heaoxiang-ai <heaoxiang@bytedance.com> Co-authored-by: Yaoyao <yuyao.yoyo@bytedance.com> Co-authored-by: kaisongli <likaisong@bytedance.com> Co-authored-by: Octopus <liyuan851277048@icloud.com> Co-authored-by: octo-patch <octo-patch@github.com> Co-authored-by: AutoCoder <wulf234@163.com> Co-authored-by: yangxinxin-7 <yangxinxin.24@bytedance.com> Co-authored-by: Yang Zhi <yangzhi.see@gmail.com> Co-authored-by: Qin Haojie <qinhaojie.exe@bytedance.com> Co-authored-by: Shawn-o <64448366+Shawn-cf-o@users.noreply.github.com> Co-authored-by: zgy <1670519171@qq.com> Co-authored-by: chuanbao666 <sunchuanbao@bytedance.com> Co-authored-by: 7. Sun <jhao.sun@gmail.com> Co-authored-by: yepper <yangshunyao@bytedance.com> Co-authored-by: mrj666 <33885009+mrj666@users.noreply.github.com> Co-authored-by: yuan7he <yuan7he@gmail.com> Co-authored-by: yuanqianhe <yuanqianhe@lattebank.com> Co-authored-by: Sense_wang <167664334+haosenwang1018@users.noreply.github.com> Co-authored-by: haosenwang1018 <haosenwang1018@users.noreply.github.com> Co-authored-by: ifeichuan <feichuan05@gmail.com> Co-authored-by: Jye10032 <736891807@qq.com>
Ferry-200
added a commit
to Ferry-200/OpenViking
that referenced
this pull request
Apr 11, 2026
* feat(encryption): add encrypt unit test for plaintext shorter than the magic header (volcengine#1217) * fix: PID lock recycle, recall threshold bypass, orphaned compressor refs (volcengine#1211) * fix: PID lock recycle, recall threshold bypass, orphaned compressor refs 1. process_lock: verify PID is actually OpenViking on Linux by checking /proc/{pid}/cmdline. Prevents false DataDirectoryLocked when PIDs are recycled to unrelated processes after crash (Fixes volcengine#1088). 2. memory-ranking: add scoreThreshold param to pickMemoriesForInjection and filter non-leaf items below threshold. Previously low-scoring memories bypassed recallScoreThreshold when supplementing leaves (Fixes volcengine#1106). 3. compressor: catch FileNotFoundError separately in _merge_into_existing, clean up orphaned vector records so they are not retried indefinitely (Fixes volcengine#1048). * style: ruff format process_lock.py --------- Co-authored-by: JasonOA888 <JasonOA888@users.noreply.github.com> * fix(queuefs): dedupe memory semantic parent enqueues within window (volcengine#769) (volcengine#792) Repeated memory writes enqueue the same parent directory for .overview/.abstract regeneration, causing redundant VLM work. Skip duplicate enqueues for the same (account, user, agent, uri) within 45s; resource/session semantic paths unchanged. Fixes volcengine#769 Made-with: Cursor * docs: add Claude Code Memory Plugin example link and Chinese docs (volcengine#1228) Add README link for Claude Code Memory Plugin example in all language variants (EN, CN, JA) and add Chinese documentation for the plugin. * fix: prevent startup hang from blocking VLM call in redo recovery (volcengine#1226) Move redo recovery to a background task so the server starts without waiting for potentially slow VLM calls. Add 60s timeout to extract_long_term_memories to prevent indefinite hangs on individual redo tasks. Clean up the redo task on stop(). Fixes volcengine#1222 Co-authored-by: Matt Van Horn <455140+mvanhorn@users.noreply.github.com> * fix(cli): handle non-UTF-8 filenames in add-resource (volcengine#1224) Replace to_string_lossy() with to_str() to preserve valid UTF-8 filenames (Chinese, Japanese, special characters) instead of silently corrupting them. Non-UTF-8 paths now return a clear Error::InvalidPath instead of garbled output. Fixes volcengine#1018 Co-authored-by: Matt Van Horn <455140+mvanhorn@users.noreply.github.com> * test(crypto): add regression tests for decrypting short plaintext files (volcengine#1223) Add tests covering the fix in PR volcengine#1163 where decrypt() raised 'Ciphertext too short' on plaintext files shorter than 4 bytes. Tests added: - test_decrypt_empty_plaintext: raw b'' returns b'' - test_decrypt_short_plaintext_less_than_4_bytes: b'X', b'AB', b'ABC' return as-is - test_decrypt_magic_prefix_without_full_header: b'OVE1' raises CorruptedCiphertextError Co-authored-by: yc111233 <yc111233@users.noreply.github.com> * Implement request-scoped wait for write APIs (volcengine#1212) fix: request wait telemetry id fix: register request wait before enqueue add log * reorg: Rewrite agfs to ragfs with rust (volcengine#1221) * reorg: rewrite agfs with rust, and named with ragfs, keep License * reorg: rewrite agfs with rust, and named with ragfs, keep License * reorg: rewrite agfs with rust, and named with ragfs, keep License * reorg: rewrite agfs with rust, and named with ragfs, keep License * reorg: rewrite agfs with rust, and named with ragfs, keep License * reorg: rewrite agfs with rust, and named with ragfs, keep License * reorg: rewrite agfs with rust, and named with ragfs, keep License * reorg: rewrite agfs with rust, and named with ragfs, keep License * fix: grep level limit * fix: grep root * fix: import error * fix: rust code optimazation * fix: CI error * fix: CI go mod cache * fix: grep level limit * fix: CI --------- Co-authored-by: openviking <openviking@example.com> * fix: update main when release, and add docker hub push (volcengine#1229) * fix security: feat(resources): harden HTTP resource ingestion against private-network SSRF (volcengine#1133) * Harden HTTP resource ingestion against private-network SSRF * chore(ci): retrigger checks * style: fix resource service import order * fix(session): add timeout to _wait_for_previous_archive_done to prevent infinite hang (volcengine#1235) When Phase 2 of a previous archive crashes after writing messages but before writing .done or .failed.json, the next archive's memory extraction enters an infinite polling loop with no exit condition. Add a 5-minute deadline. On timeout, return False (treated as failure) so the caller writes .failed.json and the session is no longer stuck. Co-authored-by: yc111233 <yc111233@users.noreply.github.com> Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com> * fix: volcengine#1238 and volcengine#1242 and volcengine#1232 (volcengine#1243) * fix(bot): respect OPENVIKING_CONFIG_FILE even when file doesn't exist Previously, bot's config path resolution would fallback to ~/.openviking/ov.conf when OPENVIKING_CONFIG_FILE was set but the file didn't exist. This was inconsistent with server's behavior, which treats a missing env-specified config as an error. In container deployments with OPENVIKING_CONFIG_FILE=/app/ov.conf, this caused bot to potentially write auto-generated config to /root/.openviking/ov.conf instead of the intended /app/ov.conf, leading to config file path mismatches. Now bot respects the environment variable unconditionally, matching server's behavior and ensuring both components use the same config path. Fixes volcengine#1242 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com> * fix(embedder): support dimension truncation for OpenAI-compatible models Previously, when users configured a dimension (e.g., 1024) for OpenAI-compatible embedding models that don't support the 'dimensions' API parameter, the system would fail with "dimensions is currently not supported" error. Additionally, when dimension was not configured, the config layer would return a hardcoded fallback of 2048, but the actual model might return a different dimension (e.g., 1024), causing dimension validation failures. This fix implements vector truncation in OpenAIDenseEmbedder: - Removes the 'dimensions' parameter from API calls (not supported by all models) - If user configures dimension=1024 and model returns 2048, truncates to 1024 - If no dimension is configured, uses model's native dimension without truncation - Applies truncation to both single and batch embedding operations This allows users to: 1. Use OpenAI-compatible models with custom dimensions via truncation 2. Control vector dimensions for storage optimization 3. Avoid dimension mismatch errors between config and actual embeddings Fixes volcengine#1238 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com> * fix: volcengine#1232 --------- Co-authored-by: openviking <openviking@example.com> Co-authored-by: Claude <noreply@anthropic.com> * Feature/memory opt (volcengine#1159) * fix(plugin): add skills to autoRecall search scope (volcengine#1225) Include viking://agent/skills in the autoRecall search alongside user memories and agent memories. Skills stored in OpenViking are now auto-injected into context when relevant to the query. Fixes volcengine#1089 Co-authored-by: Matt Van Horn <455140+mvanhorn@users.noreply.github.com> * Revert "fix(session): add timeout to _wait_for_previous_archive_done to preve…" (volcengine#1265) This reverts commit dc15daf. * channel mention (volcengine#1272) * feat(storage): volcengine vector db support sts token (volcengine#1268) * feat(storage): volcengine vector db support sts token https://www.volcengine.com/docs/7139/1302258?lang=zh * feat(storage): volcengine vector db support sts token https://www.volcengine.com/docs/7139/1302258?lang=zh * feat(server): support host all to use dual stack netstat (volcengine#1273) * feat(auth): Restrict trusted mode without API key to localhost (volcengine#1279) * Improve memory v2 lock retry controls in compressor (volcengine#1275) * fix(security): configurable embedding circuit breaker & log suppression (volcengine#1277) * update wechat group qrcode (volcengine#1282) replace the old one. * ci: optimize API test matrix from 5 to 3 channels (volcengine#1281) - Reduce OS matrix from 5 channels to 3 channels - Use ubuntu-latest, macos-latest, windows-latest instead of specific versions - Remove redundant ubuntu-arm and intel macos variants * feat: add MiniMax-M2.7 and MiniMax-M2.7-highspeed provider support (volcengine#1284) - Update provider registry to document MiniMax-M2.7 and MiniMax-M2.7-highspeed as the recommended models (replacing the outdated M2.1 example comment) - Add explicit note that MiniMax does not support system messages (they are merged into the first user message automatically) - Update README to advertise MiniMax-M2.7 and MiniMax-M2.7-highspeed as the recommended MiniMax models with configuration instructions - Add comprehensive unit tests (17 tests) covering: - Registry keyword matching for MiniMax-M2.7 and MiniMax-M2.7-highspeed - Model prefix resolution (minimax/MiniMax-M2.7) - System message merging for both LiteLLMProvider and OpenAICompatibleProvider - Edge cases: multiple system messages, no system messages, non-MiniMax models - MINIMAX_API_KEY environment variable and international API base URL Co-authored-by: octo-patch <octo-patch@github.com> * feat(eval): add openclaw eval sh (volcengine#1287) * 增加完整的一键评测脚本 * 增加完整的一键评测脚本 * add create time in add_message (volcengine#1288) * fix(lark): add lark-oapi (volcengine#1285) * docs: add prompt guides in zh and en (volcengine#1292) * benchmark: add LoCoMo evaluation scripts for mem0 (volcengine#1290) Implements a two-phase benchmark pipeline for evaluating mem0 on the LoCoMo long-term conversation dataset (10 samples, 1540 non-adversarial QA pairs). - ingest.py: imports LoCoMo conversation sessions into mem0, using sample_id as the userId namespace. All messages use "user" role with [SpeakerName]: prefix to preserve two-person dialogue structure. Temporal context is added via a [System] prefix on each session. - eval.py: sends QA questions to an OpenClaw agent backed by the openclaw-mem0 plugin. Restarts the gateway per sample to switch the active userId, verifies the correct user is loaded before running questions, then parallelizes questions within each sample using unique session keys. Parses session jsonl to collect accurate per-turn token usage. Optionally judges answers with a Volcengine ARK LLM. - delete_user.py: utility to clear mem0 memories for given user_ids. - README.md: documents prerequisites, ingest/eval parameters, output format, and per-sample run commands. * fix(docker): restore venv tooling for ragfs build (volcengine#1295) fix(docker): preserve ragfs wheel extraction script indentation fix(docker): keep heredoc cleanup inside run shell fix(docker): terminate ragfs extraction heredoc cleanly fix(ci): split docker manifest digests by registry * Include gemini optional dependency in Docker image (volcengine#1254) The Dockerfile only includes the `bot` extra when installing dependencies, which means the `google-genai` package is missing at runtime. This causes the server to crash with `TypeError: 'NoneType' object is not callable` when users configure the Gemini embedding provider. Add `--extra gemini` to the `uv sync` commands so the Docker image ships with Gemini support out of the box. Fixes volcengine#1253 * ci: add timeout and SMTP failure notification (volcengine#1293) - Add 50 minutes timeout for api_test and oc2ov_test workflows - Add SMTP email notification on workflow failure - Email will be sent to likaisong@bytedance.com Required secrets: - SMTP_USERNAME: SMTP email username - SMTP_PASSWORD: SMTP email password/app password * fix(openclaw): sanitize and cap recall queries (volcengine#1297) Recall should use cleaned user text and avoid sending oversized prompts to retrieval. Add unit coverage for sanitization and truncation behavior. * fix(ci): repair reusable build workflow yaml blocks (volcengine#1300) * feat(ast): add Lua parser support and extractor wiring (volcengine#1286) * fix(ci): remove disallowed notify-failure action (volcengine#1302) - Remove dawidd6/action-send-mail@v4 which is not in allowed list - Keep timeout-minutes: 50 configuration - Use GitHub built-in notifications instead * fix(embedder): reduce async contention in session flows (volcengine#1301) Introduce native async embedding paths across providers, switch async retrieval/session hotspots to use them, and add a standalone mixed-load benchmark plus before/after benchmark evidence for the regression. * Feat(ovpack): recursively vectorize all imported docs (volcengine#1294) * feat(ovpack): recursively enqueue directory vectorization * feat(ovpack): recursively vectorize all imported docs * refactor(ovpack): use fixed workers for direct vectorization * refactor(ovpack): revert direct vectorization to gather * chore: drop redundant default tree args * feat: 巻加测试优化功能 (volcengine#1280) - Session ID 自动管理 - 智能等待策略 - 重试机制 - 测试数据管理 * fix(eval): OpenClaw eval, import to ov use default user (volcengine#1305) * 增加完整的一键评测脚本 * 增加完整的一键评测脚本 * 完善评测脚本 * 完善评测脚本 * 完善评测脚本 * fix(memory): batch semantic processing in _process_memory_directory to prevent CPU spikes (volcengine#1245) (volcengine#1304) * Fix ci (volcengine#1307) * fix(ci): fix build ci * build: move ragfs-python packaging into setup.py * fix bug (volcengine#1317) * fix: 优化测试关键词匹配和移除 Release Approval Gate (volcengine#1313) - 添加中英文关键词匹配：删除、无、deleted、expired、no longer - 移除 workflow 中的 release-approval job * fix: remove debug print statement in bot health check endpoint (volcengine#1310) Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com> * feat: add scenario-based API tests (volcengine#1303) * feat: add scenario-based API tests - Add scenario test framework with proper categorization - Add tests for resources_retrieval, sessions, and stability_error scenarios - Add get_task() and wait_for_task() methods to API client for async operations - Add get_session_context() method for session context retrieval - Update API test workflow name from '03' to '06' - All 14 scenario tests pass with proper business logic validation * fix: skip scenarios tests when no VLM/Embedding secrets Scenarios tests require VLM for session archival summaries and Embedding for semantic search. Skip them in basic test mode. * fix(security): remove leaked token from settings.py (volcengine#1319) - Remove tests/oc2ov_test/config/settings.py containing exposed auth token - Add settings.py to .gitignore to prevent future leaks - Users should copy settings.example.py to settings.py and fill in their own tokens * fix(resources): implement trailing slash semantics for resource URIs (volcengine#1321) add support for trailing slash rules in resource URIs to control file/directory placement update CLI, API, and documentation to reflect new URI handling semantics add comprehensive tests for all URI semantics cases * Revert "fix(resources): implement trailing slash semantics for resource URIs …" (volcengine#1322) This reverts commit dec57bd. * fix litellm embedding dimension issues (volcengine#1323) Co-authored-by: openviking <openviking@example.com> * ci: optimize runner usage with conditional OS matrix and parallel limit (volcengine#1327) - Only run full OS matrix (ubuntu-24.04, macos-14, windows-latest) on main branch pushes - PR branches use ubuntu-24.04 only to reduce runner wait time - Add max-parallel: 2 to limit concurrent job execution * fix(bot):Response language, Multi user memory commit (volcengine#1329) * 修复语种问题 * 多user提交ov * ci: remove lite and full test workflows (volcengine#1331) * docs: fix docker deployment (volcengine#1332) Co-authored-by: openviking <openviking@example.com> * docs(openclaw-plugin): add health check tools guide (volcengine#1326) Add solution for health check tools report http 404 error * fix(queue): expose re-enqueue counts in queue status (volcengine#1337) Track semantic and embedding re-enqueues as first-class queue metrics so observer output, wait_processed payloads, and telemetry summaries make retry loops visible before they escalate into hard errors. * feat(s3fs): add disable_batch_delete option for OSS compatibility (volcengine#1330) (volcengine#1333) Co-authored-by: yuanqianhe <yuanqianhe@lattebank.com> * Fix/add resource cover (volcengine#1338) * fix(resources): improve handling of resource imports and naming - Add source_name to file upload requests for preserving original filenames - Handle single-directory zip files by using their root directory directly - Support viking://resources as parent directory for imports - Split summarization for resources root imports into individual child items - Add tests for new resource import behaviors * style(tests): format test files with consistent line breaks Improve readability by applying consistent line breaks in test file patches and removing trailing whitespace * afterTurn: store messages with actual roles and skip heartbeat messages (volcengine#1340) * fix: fall back to prefix filters for volcengine path scope (volcengine#1342) Co-authored-by: haosenwang1018 <haosenwang1018@users.noreply.github.com> * fix(session): auto-create missing sessions on first add (volcengine#1348) Ensure the add-message API materializes a missing session before loading it so plugins can append the first turn without an explicit create_session call. * Fix/api test issues (volcengine#1341) * fix: make api_test more robust for CI environments - Add @pytest.hookimpl(optionalhook=True) for pytest-html hooks to fix compatibility issues - test_fs_read: skip test when AGFS service is not available - test_get_overview: skip test when overview file does not exist These changes ensure tests pass gracefully on CI servers where AGFS service may not be available or files may not exist. * ci: reduce max-parallel to 1 for better resource availability Reduce max-parallel from 2 to 1 to avoid waiting for multiple runners when GitHub-hosted runners are limited. * fix: derive context_type from URI in index_resource (volcengine#1346) index_resource() always passed the default context_type="resource" to vectorize_directory_meta() and vectorize_file(), even when indexing memory directories (URIs containing /memories/). This caused all records created via the /api/v1/content/reindex endpoint to be tagged as "resource" instead of "memory", breaking stats aggregation (which filters on context_type="memory") and search scoping. Use the existing get_context_type_for_uri() helper to derive the correct context_type from the URI and pass it through to both vectorize_directory_meta() and vectorize_file(). Co-authored-by: yc111233 <yc111233@users.noreply.github.com> Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com> * reorg: remove golang depends (volcengine#1339) * docs: fix docker deployment * reorg: remove third_party/agfs * feat(s3fs): add disable_batch_delete option for OSS compatibility Port of PR volcengine#1333 from Go version to Rust: - Add disable_batch_delete config option to S3Client - When enabled, use sequential single-object deletes instead of DeleteObjects - This is for S3-compatible services like Alibaba Cloud OSS that require Content-MD5 for DeleteObjects but AWS SDK v2 does not send it by default - Add documentation and config example for OSS * fix(s3fs): pass disable_batch_delete config from Python to Rust Add disable_batch_delete to the s3_plugin_config dict in _generate_plugin_config so that the Python config can properly control the Rust S3FS plugin's behavior. * reorg: remove third_party/agfs * reorg: remove third_party/agfs * change some docs * change some docs --------- Co-authored-by: openviking <openviking@example.com> * Feat/mem opt (volcengine#1349) * fix(memory): handle string response from VLM when tools disabled - Fix AttributeError when VLM returns string instead of VLMResponse - Fix tuple creation bug with trailing comma causing double-nested tools array Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * update * chore: replace LGBTQ example with book club in entities.yaml Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(memory): disable tools on extended iteration to prevent infinite loop When max_iterations is extended due to tool calls, disable tools for the additional iteration to ensure the extract loop terminates. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(memory): handle out-of-bounds range from LLM extraction Clamp range values to valid message indices instead of skipping, to handle cases where LLM extracts incorrect ranges. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com> * fix: openai like embedding models fix, no more matryoshka error (volcengine#1350) Co-authored-by: openviking <openviking@example.com> * 增加关闭ov的配置 (volcengine#1352) * feat: 新前端项目初始化 * style: 切换到vega风格 * feat: 新增定制化的 openapi codegen * chore: 把openapi生成的ov-client移到src/gen底下 * feat: 生成 openviking-server 的 client * feat: 搭建 openviking client 的适配层 * docs: 新增 ov-client 适配层描述 * chore: 清理项目模板 * clean: 清理模板样式 * deps: 新增 tanstack query 依赖 * style: 调整 shadcn 风格 * feat: 导入基础的shadcn组件 * feat: 复刻旧的 webconsole * fix: 修复legacy代码commit不完整的问题 * docs: readme * feat: 允许不显示传递自定义client * dev: add development environment script * feat: support tags for resource management and retrieval This commit introduces a native `tags` parameter across the entire stack (Core, API, SDK, CLI) to easily tag resources and filter them during semantic search. Changes include: - **Core & Storage**: - Write `tags` to the resource root's `.meta.json` during `add_resource`. - Read tags from `.meta.json` during async semantic processing and hoist them to the VectorDB context for indexing. - Enrich directory stats/entries in `VikingFS` (`ls`, `stat`) with `tags`. - **API & Service**: Add `tags` field to resource creation and search routes. - **Python SDK**: - Add `tags` parameter to `add_resource`. - Add `tags` shortcut parameter to `find` and `search` methods, which automatically constructs the underlying `contains` metadata filters. - **Rust CLI**: Add `--tags` flag to `ov add-resource`, `ov find`, and `ov search` commands. - **Docs**: Update English and Chinese documentation for Resources, Retrieval, and Filesystem APIs to reflect the new `tags` parameters and structures. - **Tests**: Add unit tests for resource processor meta merging, VikingFS tag reading, and HTTP client tag filtering logic. * feat(search): support tags filtering and return tags in retrieval results * refactor(tags): extract build_tags_filter helper and improve validation * chore(scripts): add fast mode toggle to bootstrap menu * fix(tags): sanitize once and isolate semantic tags context * feat: 三段式布局 + 功能页拆分 - 新增顶栏(全宽) + 左侧边栏 + 右侧内容区的三段式布局 - 将 Data 大页面拆分为独立路由: FileSystem、Find、Add Memory - 将 Access 页面迁移为独立的 Settings 页面 - 侧边栏按 Data/Ops/Access 分组显示导航项 - 提取共享工具函数到 data-utils.ts - 修复 .gitignore 中 data/ 规则误匹配子目录的问题 * feat(web-studio): add i18n support and dark/light theme toggle - Add i18next with browser language detection (zh-CN / en) - Add language switcher dropdown in header bar - Add dark/light theme toggle with animated Sun/Moon icons - Wire up next-themes ThemeProvider with class-based dark mode - Replace shadcn dark theme with default Zinc (neutral gray) - Connect sidebar labels to i18n translation keys - Fix dropdown menu forced dark styling --------- Co-authored-by: baojun-zhang <zhangbaojun.1@bytedance.com> Co-authored-by: Jason <101583541+JasonOA888@users.noreply.github.com> Co-authored-by: JasonOA888 <JasonOA888@users.noreply.github.com> Co-authored-by: Protocol Zero <257158451+Protocol-zero-0@users.noreply.github.com> Co-authored-by: 灿烂甜菜 <731426007@qq.com> Co-authored-by: Matt Van Horn <mvanhorn@users.noreply.github.com> Co-authored-by: Matt Van Horn <455140+mvanhorn@users.noreply.github.com> Co-authored-by: yc111233 <109650216+yc111233@users.noreply.github.com> Co-authored-by: yc111233 <yc111233@users.noreply.github.com> Co-authored-by: Jiahui Zhou <zhoujiahui.01@bytedance.com> Co-authored-by: MaojiaSheng <shengmaojia@bytedance.com> Co-authored-by: openviking <openviking@example.com> Co-authored-by: 13ernkastel <LennonCMJ@live.com> Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com> Co-authored-by: chenjw <chenjunwen@bytedance.com> Co-authored-by: yeshion23333 <dutao.1786@bytedance.com> Co-authored-by: heaoxiang-ai <heaoxiang@bytedance.com> Co-authored-by: Yaoyao <yuyao.yoyo@bytedance.com> Co-authored-by: kaisongli <likaisong@bytedance.com> Co-authored-by: Octopus <liyuan851277048@icloud.com> Co-authored-by: octo-patch <octo-patch@github.com> Co-authored-by: AutoCoder <wulf234@163.com> Co-authored-by: yangxinxin-7 <yangxinxin.24@bytedance.com> Co-authored-by: Yang Zhi <yangzhi.see@gmail.com> Co-authored-by: Qin Haojie <qinhaojie.exe@bytedance.com> Co-authored-by: Shawn-o <64448366+Shawn-cf-o@users.noreply.github.com> Co-authored-by: zgy <1670519171@qq.com> Co-authored-by: chuanbao666 <sunchuanbao@bytedance.com> Co-authored-by: 7. Sun <jhao.sun@gmail.com> Co-authored-by: yepper <yangshunyao@bytedance.com> Co-authored-by: mrj666 <33885009+mrj666@users.noreply.github.com> Co-authored-by: yuan7he <yuan7he@gmail.com> Co-authored-by: yuanqianhe <yuanqianhe@lattebank.com> Co-authored-by: Sense_wang <167664334+haosenwang1018@users.noreply.github.com> Co-authored-by: haosenwang1018 <haosenwang1018@users.noreply.github.com> Co-authored-by: ifeichuan <feichuan05@gmail.com> Co-authored-by: Jye10032 <736891807@qq.com>
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
2 participants
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.
Description
Related Issue
Type of Change
Changes Made
Testing
Checklist
Screenshots (if applicable)
Additional Notes