Skip to content

🦞 OpenClaw Ecosystem Digest 2026-04-30 #84

@github-actions

Description

@github-actions

OpenClaw Ecosystem Digest 2026-04-30

Issues: 500 | PRs: 500 | Projects covered: 13 | Generated: 2026-04-30 02:06 UTC


OpenClaw Deep Dive

The user wants me to generate a structured project digest for OpenClaw based on the provided GitHub data. Let me analyze the data and create a comprehensive digest following the specified sections.

Let me organize the key information:

Overview:

  • 500 issues updated in last 24h (481 open, 19 closed)
  • 500 PRs updated in last 24h (469 open, 31 merged/closed)
  • 1 new release: v2026.4.27

Latest Release (v2026.4.27):

  • Codex Computer Use setup with status/install commands
  • Marketplace discovery
  • Fail-closed MCP checks for Codex-mode desktop control
  • DeepInfra joins bundled provider set with model discovery, media generation/editing, TTS, embe...

Top Issues by Comment Count:

  1. 📈 AI 开源趋势日报 2026-04-29 #75 - Linux/Windows Clawdbot Apps (101 comments, 73 👍) - Cross-platform desktop apps
  2. #12590 - memoryFlush bug (19 comments) - Memory flush reliability
  3. #72338 - Gateway CPU spin (18 comments) - Telegram stalling
  4. #22438 - Tiered bootstrap file loading (14 comments)
  5. #50090 - Community Skill Development & ClawHub (13 comments)
  6. #12602 - Slack Block Kit support (13 comments)
  7. #43735 - Skills not loading (12 comments)
  8. #73532 - Plugin loader hot loop (10 comments)
  9. #44993 - Heartbeat/Cron stale timestamp (10 comments)
  10. #18160 - Direct Exec Mode for Cron Jobs (10 comments)

Recent PRs (showing key ones):

  • #74742 - File transfer authorization fix (XL size, clawsweeper)

  • #74134 - File transfer bundled plugin (XL size)

  • #74716 - Node-hosted MCP session plumbing (XL size)

  • #46423 - Cron retry with backoff (XL size)

  • #44013 - Linux/Windows Tauri apps (L size, closed)

  • #74736 - Structured warn for bundled lock timeout (S size)

  • #74739 - Package restart verification (M size)

  • #74466 - Global skills install (S size)

  • #74744 - Signal SSE keep open (S size)

  • #74733 - Chat item timestamps (M size)

  • #74737 - Discord read-only accessor fix (XS size)

  • #74730 - Sessions spawn model override (XS size)

  • #73476 - Direct reply flag for tools (L size)

  • #74740 - Bedrock xhigh + adaptive thinking (S size)

  • #74735 - Session-scoped extra files (M size)

  • #73456 - Azure Speech realtime transcription (XL size)

  • #74719 - Handle undefined message/text (XS size)

  • #74738 - Guard openai-completions with compat flag (S size)

  • #72645 - OTel tracing (L size)

  • #74713 - Repair configured provider runtime deps (M size)

  • #69548 - Memory staggerMs option (XS size)

  • #49145 - Signal v2 reliable replies (M size)

  • #74734 - Migrate legacy ACP stream keys (M size)

Several issues stand out as particularly pressing. The memoryFlush reliability problem (#12590) appears to be a core stability concern that needs attention. Gateway CPU spinning (#72338) is causing Telegram integration failures, while the plugin loader hot loop (#73532) is creating event loop contention. There's also a structuredClone memory leak (#45438) that could be causing resource exhaustion. The Control UI webchat duplication (#71992) and Ollama thinking parameter regression (#73366) represent clear user-facing bugs that should be prioritized.

Beyond these, I'm seeing several feature gaps worth addressing. Linux and Windows desktop apps (#75) have strong community support with 73 thumbs-ups, and there's demand for Slack Block Kit integration (#12602), sensitive data masking (#64046), and MathJax rendering (#42840). The onboarding wizard (#16670) needs better memory configuration guidance, and there's appetite for backup/restore functionality (#13616), session snapshots (#13700), and session history management (#50404).

Security concerns are emerging too—Discord is leaking internal tool traces (#44905), and there's a request for built-in security scanning for skills (#45031). The exec-approvals system could benefit from denylist support (#6615), and there are multiple subagent-related issues (#44925, #50165, #47975) suggesting reliability problems in that subsystem.

OpenClaw Project Digest — 2026-04-30


1. Today's Overview

OpenClaw maintains extremely high activity with 500 issues and 500 PRs updated in the last 24 hours, indicating a project in rapid development. The release pipeline remains active with v2026.4.27 shipping major additions including Codex Computer Use infrastructure and DeepInfra as a new bundled provider. Community engagement is strong, with issue #75 (Linux/Windows desktop apps) accumulating 101 comments and 73 👍 reactions, demonstrating clear user demand for cross-platform parity. Stability concerns are emerging around memory management, gateway CPU behavior, and plugin loader performance, while the PR queue shows healthy contributions across file transfer, cron reliability, and observability features.


2. Releases

v2026.4.27 — Released 2026-04-27

Key Changes:

Component Change
Codex Computer Use New status/install commands, marketplace discovery, and fail-closed MCP checks for desktop control
DeepInfra Provider Joins bundled provider set with model discovery, media generation/editing, and TTS capabilities

Migration Notes: No breaking changes reported. DeepInfra users should verify API key configuration in openclaw.json after upgrade.


3. Project Progress

Merged/Closed PRs Today

PR Size Description Status
#44013 L Add Linux and Windows desktop apps using Tauri CLOSED
#66395 XS Fix/disable refresh during active turn CLOSED
#60027 XS Add OPENCLAW_SKIP_MODEL_WARMUP env to skip startup model pre-warm CLOSED
#42294 S Preserve audioAsVoice in outbound media sends (Telegram) CLOSED
#73094 XS Fix WS handshake ordering in ACP runtime CLOSED
#73623 S Validate plugin id before writing config CLOSED

Active PRs Advancing

PR Size Area Description
#74742 XL File Transfer Require canonical node policy authorization (ClawSweeper)
#74134 XL File Transfer New bundled plugin with file_fetch, dir_list, dir_fetch, file_write tools
#74716 XL MCP/Mac Node-hosted MCP session plumbing (first infrastructure slice for native Mac computer-use)
#46423 XL Cron Retry recurring jobs on transient errors with exponential backoff
#73456 XL Voice Add Azure Speech as realtime transcription provider for voice-call
#72645 L Observability Per-message end-to-end OTel tracing across run/skill/tool/model
#73476 L Agents Add directReply flag to tool results for bypassing LLM inference
#60868 L Agents Detect context overflow across error cause chains
#74739 M CLI Verify package restarts from updated install
#74735 M Bootstrap Support session-scoped extra files
#74733 M UI Stabilize chat item timestamps
#74734 M Config Migrate legacy ACP stream keys on load
#74713 M Plugins Repair configured provider runtime deps
#74466 S CLI Support installing skills to shared global directory via --global
#74740 S Bedrock Expose xhigh + adaptive thinking for Claude Opus 4.7
#74736 S Channels Structured warn for bundled lock timeout
#74744 S Signal Keep SSE receive stream open
#74737 XS Discord Avoid resolving tokens for read-only accessors
#74730 XS Sessions Pass model override to gateway agent method
#74719 XS UI Handle undefined message/text at runtime
#74738 S Agents Guard openai-completions tool payload with supportsTools compat flag
#69548 XS Memory Add staggerMs option to dreaming cron schedule
#49145 M Signal Reliable reply context, reaction wake, and quoted replies (v2)

4. Community Hot Topics

Most Active Issues

# Title Comments 👍 Category
#75 Linux/Windows Clawdbot Apps 101 73 Enhancement
#12590 memoryFlush does not fire reliably 19 0 Bug
#72338 Gateway CPU spin causes Telegram replies to stall 18 3 Bug
#22438 Tiered bootstrap file loading for progressive context control 14 0 Enhancement
#50090 Community Skill Development & ClawHub 13 1 Stale
#12602 Slack Block Kit support for agent messages 13 0 Enhancement
#43735 Skills not loading in agent context 12 0 Bug
#73532 Plugin loader hot loop saturates gateway 10 0 Bug
#44993 Heartbeat/Cron "Current time" timestamp is stale 10 0 Bug
#18160 Direct Exec Mode for Cron Jobs 10 9 Enhancement

Analysis

Cross-Platform Desktop Parity (#75) remains the top community request with overwhelming support (73 👍). Users want feature parity with macOS/iOS/Android apps. The Tauri-based Linux/Windows PR (#44013) was closed today—community awaits its revival.

Skill Ecosystem Gaps (#50090, #43735) indicate the ClawHub vision isn't fully realized. Skills fail to load correctly, and the community skill development pipeline needs work.

Slack Integration (#12602) shows demand for richer channel formatting beyond plain text, particularly for CRM summaries and data-heavy responses.


5. Bugs & Stability

Critical/High Severity

# Title Severity Fix PR Notes
#73532 Plugin loader hot loop saturates gateway HIGH Event loop starvation, Telegram timeouts
#72338 Gateway CPU spin causes Telegram stalls HIGH High-CPU state wedges reply pipeline
#45438 structuredClone causes ~1GB/min native memory leak HIGH RSS grows to 4-5GB in minutes
#12590 memoryFlush does not fire reliably MEDIUM Fires every other cycle only
#43747 Memory management is in chaos MEDIUM Inconsistent chunking/embedding across users

Medium Severity

# Title Fix PR Notes
#71992 Control UI webchat duplicates every assistant reply Regression from #5964/#39469
#73366 Ollama thinking parameter always forced to false Regression after 2026.4.26
#43735 Skills not loading from ~/.openclaw/workspace/skills/ Subset only appears in context
#40540 openclaw update fails with EBUSY on Windows Prevents self-update
#39038 Windows 11 24H2 node startup hangs on PATH Crash/hang bug

Lower Severity (Notable Patterns)

# Title Notes
#44905 Discord leaks internal tool-call traces Security concern
#44925 Subagent completion silently lost No retry/notification
#46637 reasoning_content causes JSON parse error Qwen 3.5 compatibility
#45269 apply_patch treated as unknown tool Regression

6. Feature Requests & Roadmap Signals

High-Impact Requests (by 👍 or comment velocity)

# Title 👍 Signals
#75 Linux/Windows Clawdbot Apps 73 Cross-platform parity demand
#18160 Direct Exec Mode for Cron Jobs 9 Performance/timeout concerns
#42840 Add MathJax/LaTeX Support to Control UI 4 Scientific/technical user base
#6615 Add denylist support for exec-approvals 6 Security/flexibility
#50880 Steer queue mode silently degrades 2 Reliability concern
#48814 Pre-send queue check for stale replies 2 UX refinement

Predicted Next-Version Candidates

Based on PR activity and issue engagement:

  1. File Transfer Plugin (#74134, #74742) — XL-sized PRs indicate imminent merge
  2. Cron Retry with Backoff (#46423) — Large, well-developed PR addresses reliability
  3. Node-hosted MCP Session (#74716) — First infrastructure slice for Mac computer-use
  4. Azure Speech Transcription (#73456) — Expands voice-call provider options
  5. OTel End-to-End Tracing (#72645) — Addresses observability gap in regulated environments

Ecosystem Expansion Signals

Request Issue Need
Global skills install #74466 (PR) Multi-agent shared capabilities
Sensitive data masking #64046 Enterprise security
Backup/restore utility #13616 DevOps/migration
Session snapshots #13700 Workflow flexibility
Security scanning for skills #45031 Trust/safety

7. User Feedback Summary

Pain Points

Category Evidence Impact
Memory Reliability #12590, #43747, #45438 Users report inconsistent behavior; memory flushes fail, chunking varies by user, native memory leaks
Gateway Stability #72338, #73532 High CPU states cause Telegram stalling and event loop starvation
Subagent Reliability #44925, #50165, #47975 Results silently lost, sessions persist after completion, completion announced prematurely
Cross-Platform Gaps #75, #40540, #39038

Cross-Ecosystem Comparison

The user wants a cross-project comparison report based on all the project digests I've generated. Let me analyze the data from each project and create a comprehensive comparison.

Let me gather the key metrics from each project:

  1. OpenClaw (github.com/openclaw/openclaw)

    • Issues: 500 updated in last 24h (481 open, 19 closed)
    • PRs: 500 updated in last 24h (469 open, 31 merged/closed)
    • Releases: 1 new (v2026.4.27)
    • Health: Very high activity, core reference project
  2. NanoBot (github.com/HKUDS/nanobot)

    • Issues: 13 updated (4 open, 9 closed)
    • PRs: 38 updated (12 open, 26 merged/closed)
    • Releases: 1 new (v0.1.5.post3)
    • Health: High activity, strong community engagement
  3. Zeroclaw (github.com/zeroclaw-labs/zeroclaw)

    • Issues: 50 updated (30 open, 20 closed)
    • PRs: 50 updated (49 open, 1 merged/closed)
    • Releases: 0
    • Health: High activity but low merge rate
  4. PicoClaw (github.com/sipeed/picoclaw)

    • Issues: 13 (12 open, 1 closed)
    • PRs: 20 (15 open, 5 merged/closed)
    • Releases: 1 nightly (v0.2.7-nightly.20260430.a36472b5)
    • Health: High PR activity, active development
  5. NanoClaw (github.com/qwibitai/nanoclaw)

    • Issues: 3 (2 open, 1 closed)
    • PRs: 50 (29 open, 21 merged/closed)
    • Releases: 0
    • Health: Very high PR activity
  6. IronClaw (github.com/nearai/ironclaw)

    • Issues: 26 (25 open, 1 closed)
    • PRs: 50 (18 open, 32 merged/closed)
    • Releases: 1 (ironclaw-v0.27.0)
    • Health: High activity, major architecture (Reborn) in progress
  7. LobsterAI (github.com/netease-youdao/LobsterAI)

    • Issues: 1 (1 open, 0 closed)
    • PRs: 28 (24 open, 4 merged/closed)
    • Releases: 1 (2026.4.29)
    • Health: Moderate activity with strong release cadence
  8. TinyClaw (github.com/TinyAGI/tinyclaw)

    • No activity in last 24h
  9. Mol tis (github.com/moltis-org/moltis)

    • Issues: 6 (4 open, 2 closed)
    • PRs: 8 (5 open, 3 merged/closed)
    • Releases: 2 (20260429.02, 20260429.01)
    • Health: Moderate activity, steady releases
  10. CoPaw (github.com/agentscope-ai/CoPaw)

    • Issues: 29 (15 open, 14 closed)
    • PRs: 20 (9 open, 11 merged/closed)
    • Releases: 1 (v1.1.5)
    • Health: High activity, good merge rate
  11. ZeptoClaw (github.com/qhkm/zeptoclaw)

    • No activity in last 24h
  12. EasyClaw (github.com/gaoyangz77/easyclaw)

    • No activity in last 24h
  13. NemoClaw (github.com/NVIDIA/NemoClaw)

    • Issues: 50 (36 open, 14 closed)
    • PRs: 50 (31 open, 19 merged/closed)
    • Releases: 0
    • Health: High activity, strong PR engagement

Now I'll organize this into a cross-project comparison report with the requested sections.

Personal AI Assistant & Agent Open-Source Ecosystem — Cross-Project Comparison Report

Date: 2026-04-30 | Scope: 13 tracked projects


1. Ecosystem Overview

The personal AI assistant and agent open-source ecosystem is in a high-velocity growth phase, with the majority of tracked projects showing daily activity across issues and pull requests. OpenClaw, as the core reference implementation, anchors the ecosystem with the highest activity volume (500 issues + 500 PRs per day), while a cohort of derivative and companion projects—NanoBot, Zeroclaw, PicoClaw, NanoClaw, IronClaw, and NemoClaw—each maintain substantial momentum. The ecosystem collectively addresses multi-channel integration (Telegram, Slack, WeChat, DingTalk, Feishu, WhatsApp), multi-provider LLM support, skill/plugin architectures, and security hardening. A notable trend is the emergence of Rust-based rewrites and Reborn-style architectural overhauls (IronClaw), signaling a maturation from prototype to production-grade infrastructure. Two projects (TinyClaw, ZeptoClaw, EasyClaw) show no recent activity, suggesting either abandonment or a shift to private development.


2. Activity Comparison

Project Issues (24h) PRs (24h) Releases (24h) Open Issues Open PRs Merge Rate Health Tier
OpenClaw 500 500 1 (v2026.4.27) 481 469 ~6% 🥇 Tier 1
NanoBot 13 38 1 (v0.1.5.post3) 4 12 ~68% 🥈 Tier 2
Zeroclaw 50 50 0 30 49 ~2% 🥈 Tier 2
PicoClaw 13 20 1 (nightly) 12 15 ~25% 🥈 Tier 2
NanoClaw 3 50 0 2 29 ~42% 🥈 Tier 2
IronClaw 26 50 1 (v0.27.0) 25 18 ~64% 🥈 Tier 2
LobsterAI 1 28 1 (2026.4.29) 1 24 ~14% 🥉 Tier 3
Moltis 6 8 2 (20260429.01/02) 4 5 ~38% 🥉 Tier 3
CoPaw 29 20 1 (v1.1.5) 15 9 ~55% 🥈 Tier 2
NemoClaw 50 50 0 36 31 ~38% 🥈 Tier 2
TinyClaw 0 0 0 ⚠️ Inactive
ZeptoClaw 0 0 0 ⚠️ Inactive
EasyClaw 0 0 0 ⚠️ Inactive

Notes:

  • Merge Rate = merged/closed PRs ÷ total PRs updated in 24h. Higher is better for throughput.
  • Health Tier 1 = core ecosystem anchor with highest volume. Tier 2 = active, healthy development. Tier 3 = moderate activity, improving. Inactive = no updates in 24h.
  • OpenClaw's low merge rate (6%) reflects its role as a merge queue for a large contributor base, not a health concern.

3. OpenClaw's Position

Advantages vs. Peers

Dimension OpenClaw Peers
Activity Volume 500 issues + 500 PRs/day Next highest: 50/50 (Zeroclaw, NemoClaw) — 10× lower
Contributor Base 12 new contributors in NanoBot; IronClaw has multiple contributors Smaller, more concentrated teams
Release Cadence Weekly releases (v2026.4.27) Variable; some projects have infrequent releases
Ecosystem Role Core reference; all other Claw projects are derivatives or companions Dependent on OpenClaw for core agent logic
Feature Breadth MCP, multi-channel, multi-provider, Codex Computer Use, skill system Narrower scope, often focused on specific channels or providers

Technical Approach Differences

  • OpenClaw (TypeScript) serves as the canonical implementation, with Zeroclaw (Rust) and PicoClaw (Go) providing alternative language runtimes for different deployment contexts.
  • IronClaw's Reborn architecture represents the most ambitious technical divergence—a ground-up Rust rewrite with WASM tool runtimes, capability contracts, and a service-layer decomposition that OpenClaw does not yet match in structure.
  • NanoBot and CoPaw focus on Chinese-market channels (Feishu, DingTalk, WeChat, QQ), filling a geographic gap that OpenClaw's Western-centric channel set does not cover.
  • NemoClaw targets NVIDIA DGX/Spark infrastructure, positioning itself as an enterprise GPU-cluster deployment option.

Community Size Comparison

OpenClaw's community is the largest by raw volume. However, NanoBot's v0.1.5.post3 release attracted 12 new contributors in a single cycle, suggesting aggressive community growth. IronClaw's 38-comment epic issue (#2987) indicates strong internal coordination but a smaller external contributor base.


4. Shared Technical Focus Areas

The following requirements are emerging across multiple projects simultaneously, indicating ecosystem-wide demand:

Multi-Channel Parity

Projects: OpenClaw, Zeroclaw, PicoClaw, NanoBot, CoPaw, NemoClaw
Need: Consistent behavior across Telegram, Slack, WeChat, DingTalk, Feishu, WhatsApp, Matrix, and Discord. Specific pain points include:

  • File/attachment sending (CoPaw #981, LobsterAI #1877)
  • Voice message handling (PicoClaw #2704, CoPaw #3887)
  • Thread/reply context isolation (NanoBot #3543, Zeroclaw #6226)

Provider Diversification

Projects: OpenClaw, Zeroclaw, PicoClaw, NanoClaw, LobsterAI, CoPaw
Need: Support for more LLM providers beyond OpenAI/Anthropic. Requests include:

  • Google Gemini (NanoClaw #2136, CoPaw #3940)
  • DeepSeek with thinking/reasoning modes (PicoClaw #2706, Zeroclaw #5289)
  • Local inference (Ollama, vLLM, Intel OpenVINO) (Zeroclaw #5459, PicoClaw #2703)
  • Bedrock streaming (PicoClaw #2645, OpenClaw #74740)

Memory & Context Management

Projects: OpenClaw, Zeroclaw, NanoClaw, CoPaw, NemoClaw
Need: Reliable memory persistence, session continuity, and context compaction. Specific issues:

  • Memory flush reliability (OpenClaw #12590)
  • Session ID mismatches (Zeroclaw #5550)
  • Context compaction thresholds (NanoClaw #2109, CoPaw #3944)
  • Memory search for CJK text (CoPaw #3811)

Security Hardening

Projects: OpenClaw, Zeroclaw, PicoClaw, Moltis, CoPaw, NemoClaw
Need: Sandboxing, path validation, credential isolation. Specific needs:

  • Sandbox escape prevention (Moltis #923, OpenClaw #73532)
  • Path traversal guards (Zeroclaw #5518, PicoClaw #1042)
  • Credential injection boundaries (IronClaw #3068, CoPaw #3955)
  • URL protocol whitelisting (LobsterAI #869, #877)

Skill/Plugin Architecture

Projects: OpenClaw, NanoBot, PicoClaw, CoPaw, LobsterAI
Need: Standardized skill distribution, installation, and runtime. Specific needs:

  • Global/shared skill installation (OpenClaw #74466)
  • Skill compilation for token minimization (Zeroclaw #5146)
  • Skill marketplace/registry (OpenClaw #50090)
  • MCP tool integration (NanoBot #3541, PicoClaw #2696)

Observability & Debugging

Projects: OpenClaw, Zeroclaw, IronClaw, PicoClaw, CoPaw
Need: Tracing, logging, and debugging tools for production deployments.

  • OTel end-to-end tracing (OpenClaw #72645, Zeroclaw #6009)
  • Canary lane monitoring (IronClaw #3075, #3074)
  • Session snapshot/export (Zeroclaw #13700, LobsterAI #853)

5. Differentiation Analysis

Project Primary Language Target User Architecture Focus Key Differentiator
OpenClaw TypeScript General users, developers Monolithic agent with plugin system Core reference; broadest channel/provider support
Zeroclaw Rust Performance-critical deployments Modular Rust agent Memory safety, performance, structured concurrency
PicoClaw Go Resource-constrained environments Lightweight CLI-first agent Small binary, MQTT/WhatsApp/Intel OpenVINO support
NanoBot Chinese enterprise users Conversation-first architecture Feishu group isolation, HookCenter plugin discovery
NanoClaw Developers, power users Container-based agent Docker sandbox, knowledge base compilation, MCP support
IronClaw Rust Enterprise/regulated environments Reborn service-layer architecture WASM tool runtimes, capability contracts, OTel-native
LobsterAI Chinese developers Desktop app (Electron/Tauri) ChatGPT OAuth, Xiaomi/Baidu providers, spec documentation
Moltis Developers Modular agent Voice personas, sandbox security, telephony (Twilio)
CoPaw Chinese enterprise teams Multi-agent orchestration Agent workspace isolation, inter-agent task progress
NemoClaw NVIDIA DGX/Spark users GPU cluster deployment Containerized sandbox, PVC persistence, Brev integration

Key Architectural Divergences:

  • IronClaw vs. OpenClaw: IronClaw's Reborn architecture decomposes the agent into discrete services (AgentLoop, ToolSurface, SessionThread, CapabilityHost) with WASM-based tool execution. OpenClaw remains monolithic, trading modularity for simplicity.
  • PicoClaw vs. OpenClaw: PicoClaw targets embedded/CLI use cases with a minimal footprint, while OpenClaw targets full-featured server deployments.
  • CoPaw vs. OpenClaw: CoPaw emphasizes multi-agent orchestration with workspace isolation, whereas OpenClaw focuses on single-agent multi-channel scenarios.
  • NemoClaw vs. OpenClaw: NemoClaw is purpose-built for GPU cluster orchestration with Kubernetes-native features, while OpenClaw is platform-agnostic.

6. Community Momentum & Maturity

Activity Tiers

Tier Projects Characteristics
Rapid Iteration OpenClaw, NanoClaw, Zeroclaw, NemoClaw 50+ PRs/day; large open PR backlogs; weekly releases; high bug velocity
Healthy Growth NanoBot, IronClaw, CoPaw, PicoClaw 13–38 PRs/day; good merge rates (25–68%); active feature development
Steady Development Moltis, LobsterAI 6–28 PRs/day; moderate merge rates; focused feature delivery
Stalled/Inactive TinyClaw, ZeptoClaw, EasyClaw No activity in 24h; may be archived or in private development

Rapidly Iterating Projects

  • OpenClaw: The 500/500 daily activity is unsustainable for maintainers long-term. The low merge rate (6%) suggests a bottleneck. The project needs more reviewers or automated tooling to prevent PR backlog growth.
  • NanoClaw: 50 PRs with 42% merge rate is impressive for a smaller project. The rapid feature additions (Gemini provider, knowledge base, remote MCP) suggest a small but highly productive team.
  • Zeroclaw: 50 PRs but only 2% merge rate is a warning sign. The project may be generating more PRs than maintainers can review, or PRs are being held for larger batch merges.
  • NemoClaw: High activity with 38% merge rate indicates healthy throughput. The Docker build failures (#2689, #2686) suggest the project needs more CI coverage.

Stabilizing Projects

  • IronClaw: The Reborn architecture landing (v0.27.0) represents a major inflection point. Post-Reborn, the project may shift from rapid architectural change to feature refinement and stability hardening.
  • CoPaw: Good merge rate (55%) and balanced bug/feature work suggest a maturing project. The security vulnerability (#3955) closed same-day indicates responsive maintenance.

7. Trend Signals

The following signals extracted from community discussions and issue/PR activity are relevant for AI agent developers and decision-makers:

1. Multi-Modal Input is a Hard Requirement

Evidence: PicoClaw image handling bugs (#2718), CoPaw multimodal message support (PR #3509), NanoClaw image upload regression (#2139), OpenClaw Codex Computer Use for desktop control.
Implication: Agents must handle images, audio, and video natively across all channels. Providers without multimodal support (e.g., DeepSeek with image_url errors) need graceful fallback.

2. Local/Inference Infrastructure is a Growing Segment

Evidence: PicoClaw Intel OpenVINO support (#2703), NemoClaw DGX/Spark focus, Zeroclaw Ollama provider bugs (#5459


Peer Project Reports

NanoBotHKUDS/nanobot

Let me analyze the GitHub data for NanoBot and create a structured project digest for 2026-04-30.

Key data points:

  • Issues updated in last 24h: 13 (open/active: 4, closed: 9)
  • PRs updated in last 24h: 38 (open: 12, merged/closed: 26)
  • New releases: 1 (v0.1.5.post3)

Let me organize this into the required sections.

  1. Today's Overview: High activity day with 38 PRs and 13 issues updated. New release v0.1.5.post3 with 57 PRs merged and 12 new contributors. Focus on conversation/threading features.

  2. Releases: v0.1.5.post3 - conversations becoming first-class citizens, Feishu group topics get isolated sessions, 57 PRs merged, 12 new contributors.

  3. Project Progress: 26 PRs merged/closed. Key merged PRs include:

    • HookCenter infrastructure
    • Per-channel progress controls
    • Atomic writes for history.jsonl
    • Olostep web search provider
    • create-instance skill
    • Matrix user_id sanitization
    • WeChat context_token refresh fix
  4. Community Hot Topics:

    • Issue #3095: custom provider with anthropic api (6 comments)
    • Issue #2590: minimax provider broken after upgrade (5 comments)
    • Issue #1099: cron request param config (3 comments)
    • PR #3542: multi-account support for WeChat (open)
    • PR #3543: fix for reply_in_thread in Feishu groups (open)
  5. Bugs & Stability:

    • Issue #3533: reply_in_thread forced in group chats (bug, PR #3543 fixes)
    • Issue #2590: minimax provider regression
    • Issue #1783: Codex provider hardcoded 60s timeout
    • Issue #2341: WebFetchTool proxies through Jina regardless of API key
  6. Feature Requests & Roadmap Signals:

    • Issue #3518: xiaomi models support
    • PR #3542: multi-account WeChat support
    • PR #3538: gateway start/stop/restart commands
    • PR #3539: upgrade wizard skill
    • PR #3534: CLAUDE.md for AI contributors
  7. User Feedback Summary: Users reporting issues with local models hallucinating, feedback on unreleased version being too many questions and incapable, cron job connection pool issues.

  8. Backlog Watch: Issue #1068: Local hallucinating (stale but active), Issue #877: Feedback on unreleased version, Issue #2590: minimax provider regression still open.

Now I'll format this into the structured digest.

NanoBot Project Digest — 2026-04-30

1. Today's Overview

NanoBot experienced a highly productive day with 38 PRs updated and 13 issues processed. The project released v0.1.5.post3, marking conversations as first-class citizens with 57 PRs merged and 12 new contributors joining. Channel-specific improvements dominated the day, particularly around Feishu threading behavior, WeChat multi-account support, and per-channel configuration flexibility. The community is actively engaged with multiple feature proposals and bug reports, indicating healthy project momentum heading into the next release cycle.


2. Releases

v0.1.5.post3 — Conversations as First-Class Citizens

Released: 2026-04-29 | PRs Merged: 57 | New Contributors: 12

This release shifts focus from reach and polish to conversation-first architecture. Key highlights:

  • Feishu group topics now receive isolated sessions, enabling cleaner multi-thread interactions
  • 57 PRs merged including threading fixes, provider improvements, and infrastructure enhancements
  • 12 new contributors welcomed to the project

"If v0.1.5.post2 was about reach and polish, v0.1.5.post3 is about conversations becoming first-class citizens of their platform."

Migration Notes: No breaking changes reported. Users of Feishu group chats may notice altered threading behavior as intended.


3. Project Progress

Merged/Closed PRs (26 total)

PR Author Description
#3541 @aiguozhi123456 HookCenter infrastructure — centralized hook points with plugin discovery via entry_points
#3487 @boogieLing Per-channel progress controlssendProgress/sendToolHints configurable per channel
#3508 @yorkhellen Atomic writes for history.jsonl — prevents corruption on crash/power loss
#3505 @chengyongru Olostep web search provider — backported from nightly
#3457 @chengyongru create-instance skill — create new bot instances via conversation
#3510 @JiajunBernoulli Matrix user_id sanitization — fixes Windows file path issues
#3517 @XJPeng12 WeChat context_token refresh — fixes silent message drops in cron jobs
#3530 @x804907 Agent workflow execution — 6-stage structured workflow with env toggle
#3529 @x804907 Security enhancements — path traversal detection, sensitive data sanitization

Open PRs (12 total)

PR Author Description
#3542 @gracece WeChat multi-account support — run multiple personal accounts simultaneously
#3543 @04cb Fix Feishu reply_in_thread — respects replyToMessage config in groups
#3538 @choiking Gateway start/stop/restart commands — CLI management for gateway
#3539 @chengyongru Upgrade wizard skill — two-skill system for bot updates
#3534 @chengyongru CLAUDE.md for AI contributors — guides for AI coding assistants
#3532 @boogieLing Respect max_iterations for subagents — replaces hardcoded limit of 15

4. Community Hot Topics

Most Active Issues (by comments)

  1. #3095 — Custom provider with Anthropic API (6 comments, Closed)

    • Topic: Using custom endpoints that implement Anthropic API but aren't Anthropic's official endpoint
    • Signal: Users need flexibility in provider endpoint configuration beyond official providers
  2. #2590 — MiniMax provider broken after v0.1.4.post6 (5 comments, Open)

    • Topic: Built-in MiniMax provider stopped working after upgrade
    • Signal: Regression affecting production deployments; needs priority fix
  3. #1099 — Cron request param configuration (3 comments, Closed)

    • Topic: Connection pool exhaustion in Telegram cron jobs
    • Signal: Need for configurable connection pool sizing for high-frequency cron tasks

Most Active PRs (by attention)

  • #3542 — WeChat multi-account support (Open)

    • High community interest in running multiple WeChat accounts from single instance
    • Backward compatible design makes this a low-risk enhancement
  • #3543 — Feishu reply_in_thread fix (Open)

    • Directly addresses bug #3533; routing through shared _thread_reply_target helper
    • Expected to merge soon given clear problem/solution

5. Bugs & Stability

Critical/High Severity

Issue Description Status Fix PR
#2590 MiniMax provider regression — broken since v0.1.4.post6 Open None
#1783 Codex hardcoded 60s timeout — causes silent failures, no retry Closed None
#2341 WebFetchTool proxies through Jina regardless of JINA_API_KEY — privacy concern Closed None

Medium Severity

Issue Description Status Fix PR
#3533 reply_in_thread forced in Feishu groups — ignores replyToMessage config Open #3543
#3517 WeChat cron messages silently dropped — missing context_token Merged

Notable Stability Fixes Merged Today

  • #3508: Atomic writes for history.jsonl using temp file + os.replace() + fsync() pattern
  • #3510: Matrix channel Windows compatibility fix for colon in user_id

6. Feature Requests & Roadmap Signals

User-Requested Features

Issue/PR Request Priority Signal
#3518 Xiaomi models support New request, 2 comments
#3452 Per-channel sendProgress/sendToolHints Implemented in #3487
#3542 WeChat multi-account support Open, high community interest
#3538 Gateway start/stop/restart commands Open, CLI enhancement

Roadmap Indicators

  1. Multi-channel flexibility — Per-channel configuration (progress controls, reply behavior) is a clear trend
  2. Provider expansion — Requests for Xiaomi models suggest demand for more provider options
  3. Operational tooling — Gateway management commands indicate need for better DevOps integration
  4. Agent workflow — Multiple PRs (#3530, #3531, #3535) exploring structured agent execution workflows

7. User Feedback Summary

Pain Points

Feedback Source Issue
Local models hallucinating @Claudioappassionato #1068 — persists across all tested models, likely configuration issue
Unreleased version "too many questions and incapable" @kkayachr #877 — feedback on latest commit behavior
MiniMax provider broken @pmchan #2590 — regression blocking production use
Cron connection pool exhaustion @zwbdzb #1099 — Telegram cron jobs hitting limits

Positive Signals

  • v0.1.5.post3 release received enthusiastic response with 57 merged PRs and 12 new contributors
  • HookCenter infrastructure (#3541) enables better plugin ecosystem
  • create-instance skill (#3457) praised for enabling bot creation via conversation

8. Backlog Watch

Long-Unanswered Important Issues

Issue Age Status Concern
#1068 — Local hallucinating ~2 months Open (stale) Affects all local models; likely systemic config issue
#877 — Unreleased version feedback ~2 months Open User experience concerns with latest commits
#2590 — MiniMax regression ~1 month Open Blocking production deployments

PRs Needing Maintainer Attention

PR Age Status Concern
#2867 — Telegram group allowlist ~24 days Open (invalid flagged) Marked invalid but may need review
#3542 — WeChat multi-account 1 day Open High community interest, ready for review

Digest generated from NanoBot (HKUDS/nanobot) GitHub data for 2026-04-30

Zeroclawzeroclaw-labs/zeroclaw

Let me analyze the provided GitHub data for Zeroclaw and generate a structured project digest for 2026-04-30.

Key data points:

  • Issues updated in last 24h: 50 (open/active: 30, closed: 20)
  • PRs updated in last 24h: 50 (open: 49, merged/closed: 1)
  • New releases: 0

Let me organize the information:

Latest Releases: None

Latest Issues (showing top 30 by comment count):

  1. #6123 [OPEN] - default_model issue on fresh install - 15 comments, S1 severity

  2. #5509 [CLOSED] - Telegram voice message transcription support - 7 comments

  3. #5146 [OPEN] - Token consumption minimization via skill compilation - 6 comments, high risk

  4. #5550 [OPEN] - memory recall session_id mismatch bug - 6 comments, medium risk

  5. #6153 [OPEN] - Matrix voice transcription failed - 5 comments, S2 severity

  6. #5459 [CLOSED] - Ollama provider tool_count=0 bug - 5 comments

  7. #5125 [CLOSED] - CPU spikes when typing - 4 comments, S0 severity

  8. #5244 [OPEN] - Dashboard Channels tab crash - 4 comments, S2 severity

  9. #5501 [OPEN] - trigger cron manually - 3 comments

  10. #5518 [OPEN] - forbidden_path_argument blocks safe redirect targets - 3 comments, S0 severity

  11. #5170 [OPEN] - memory_recall with query '*' returns empty - 3 comments

  12. #5475 [OPEN] - Copilot + Telegram Invalid parameter - 3 comments, S1 severity

  13. #5470 [OPEN] - Multiple issues when running safely - 3 comments, S2 severity

  14. #5415 [OPEN] - Context spillage from chat to schedule - 3 comments, S0 severity

  15. #5360 [OPEN] - codex_cli tool passes unsupported -q flag - 3 comments

  16. #6090 [CLOSED] - Telegram misconfigured anthropic call - 3 comments, S0 severity

  17. #5553 [CLOSED] - Shell tool executes but Telegram returns raw tool_calls JSON - 3 comments, S1 severity

  18. #5502 [CLOSED] - Add allowed_tools configuration option - 3 comments

  19. #6210 [OPEN] - SkillForge auto-integrator emits non-schema fields - 2 comments

  20. #5356 [OPEN] - Canvas tool from channel server writes to separate CanvasStore - 2 comments

  21. #6128 [OPEN] - add deny_unknown_fields to SkillMeta - 2 comments

  22. #6073 [CLOSED] - Web UI Settings editing broken - 2 comments, S3 severity

  23. #6094 [CLOSED] - Dead duplicate of cost/tracker.rs - 2 comments, S3 severity

  24. #5289 [OPEN] - Bedrock provider sends API_KEY as x-api-key header - 2 comments, P1 priority

  25. #5861 [CLOSED] - voice-wake feature fails to compile - 2 comments, S1 severity

  26. #5503 [CLOSED] - Pre-LLM command shortcuts - 2 comments

  27. #5498 [CLOSED] - fix release pipeline - 2 comments

  28. #6226 [CLOSED] - Slack thread_replies=true synthesises thread_ts - 1 comment, S2 severity

  29. #6225 [OPEN] - Smart Truncation for Telegram - 1 comment

  30. #5701 [CLOSED] - add clear_messages to SessionBackend trait - 1 comment

The pull requests show significant activity with one closure and several open items. Notable work includes a large-scale enhancement to web onboarding via per-property CRUD endpoints, a cron fix enabling WhatsApp as a delivery channel, and a bug fix addressing tool output formatting in channels. Several enhancements are in progress covering web interface improvements, ACP protocol implementation, session key sanitization, and chat input functionality.

Additional pull requests address installation fixes for web dashboard extraction, codex CLI compatibility, webhook retry logic with exponential backoff, Docker sandbox workspace mounting, git command security distinctions, rustls initialization, skill TOML parsing, voice capture buffering, OpenTelemetry span enrichment, and provider-level fixes for media markers and Gemini quota handling.

The project shows substantial activity with 50 issues and pull requests updated in the past day, though only one PR has merged. Several critical issues are emerging: a default_model configuration problem on fresh installs, token consumption optimization through skill compilation, memory recall session mismatches, and Matrix voice transcription failures. Security concerns include forbidden_path_argument blocking legitimate redirect targets and context spillage between chat and scheduled contexts. The Ollama provider has a tool_count bug, and the dashboard's Channels tab is crashing. Recent work includes Telegram voice transcription support, CPU spike fixes, and an allowed_tools configuration option.

Zeroclaw Project Digest — 2026-04-30

1. Today's Overview

Zeroclaw shows high activity on April 30, 2026, with 50 issues and 50 PRs updated in the last 24 hours. The project has 30 open/active issues and 49 open PRs, indicating substantial development momentum. No new releases were published today. Notable attention is focused on multi-channel integration bugs (Telegram, Matrix, Slack), provider configuration issues (Ollama, Bedrock), and security-related concerns around path validation and context isolation. The community is actively contributing both bug fixes and enhancements, with several high-risk items demanding immediate maintainer review.


2. Releases

No new releases today. The project last shipped v0.6.9 (see issue #5498 for release pipeline fixes). Users are advised to monitor the releases page for upcoming versions addressing the high-risk bugs currently under active development.


3. Project Progress

Merged/Closed PRs Today

PR Description Status
#6035 fix(channels/acp): correct tool output formatting CLOSED
#5498 fix: configure missing secrets/permissions for release pipeline CLOSED

Active PRs Advancing

PR Description Risk
#6179 feat(gateway,web,cli): web onboarding parity via per-property CRUD endpoints HIGH
#6230 fix(cron): allow whatsapp as cron delivery channel LOW
#6167 fix(channels/acp): implement ACP protocol v1 with tool-call permission and back-channel HIGH
#5978 feat(voice): speech capture buffer + STT dispatch HIGH
#5905 fix: add workspace bind-mount support to DockerSandbox HIGH
#5838 feat(channels/webhook): add retry logic with exponential backoff MEDIUM

4. Community Hot Topics

Most Active Issues (by comment count)

Issue Title Comments Status
#6123 default_model issue on fresh install 15 OPEN
#5509 Telegram voice message transcription support 7 CLOSED
#5146 Token consumption minimization via skill compilation 6 OPEN
#5550 memory recall session_id mismatch bug 6 OPEN

Analysis

  • #6123 (15 comments): Fresh LXC container installs fail with default_model error during onboarding. This S1 workflow blocker indicates onboarding UX needs hardening for containerized deployments.
  • #5509 (7 comments, now closed): Feature request for Telegram voice transcription was addressed, closing a parity gap with OpenClaw (TypeScript counterpart).
  • #5146 (6 comments): High-risk enhancement request for skill compilation to minimize token usage. Users report 400+ line SKILL.md files being sent for simple queries like "what's the weather."
  • #5550 (6 comments): Autosaved Conversation memories invisible due to session_id mismatch between store and recall paths.

Most Active PRs

  • #6179: Large-scale web onboarding parity PR (size: XL) introducing per-property CRUD endpoints under /api/config/*.
  • #5978: Voice capture buffer and STT dispatch feature (size: XL) addressing speech handling gaps.

5. Bugs & Stability

Critical/High Severity (S0-S1)

Issue Title Severity Fix PR?
#6123 default_model issue on fresh install S1 No
#5518 forbidden_path_argument blocks safe redirect targets (/dev/null, /dev/stdout) S0 No
#5415 Context spillage from chat to schedule S0 No
#5475 Copilot + Telegram result in Invalid parameter S1 No
#5459 Ollama provider sends tool_count=0 — native tool calling broken S1 No
#5553 Shell tool executes but Telegram returns raw tool_calls JSON S1 No

Medium Severity (S2)

Issue Title Status
#6153 Matrix: voice transcription failed: Unsupported audio format OPEN
#5244 Dashboard: Channels tab crash and Overview render error on v0.6.8 OPEN
#5470 Multiple issues when running safely OPEN
#6226 Slack thread_replies=true synthesises thread_ts breaking session continuity CLOSED

Notable Fixes Merged

  • #6035: ACP tool output formatting corrected
  • #5498: Release pipeline secrets/permissions configured

6. Feature Requests & Roadmap Signals

High-Impact Feature Requests

Issue Title Priority Signal
#5146 Token consumption minimization via skill compilation HIGH (6 comments, 👍1)
#5501 trigger cron manually Enhancement (3 comments)
#6225 Smart Truncation for Telegram Enhancement (1 comment)
#5502 Add allowed_tools configuration option to AgentConfig CLOSED (merged)

Roadmap Signals

  • Voice/Speech: Active development in #5978 for speech capture buffering and STT dispatch
  • Multi-channel parity: Telegram voice transcription (#5509) closed; WhatsApp cron delivery (#6230) in review
  • Security hardening: Path validation (#5518), git command distinction (#5939), workspace isolation (#5905)
  • Observability: OTel tool span enrichment (#6009) in progress

7. User Feedback Summary

Pain Points

  1. Onboarding failures: Fresh installs on LXC containers hit default_model errors immediately (#6123)
  2. Ollama tool calling broken: Users with self-hosted Ollama cannot use native tool calling due to hardcoded tool_count=0 (#5459)
  3. Memory recall failures: Session ID mismatches cause silent data loss in memory system (#5550)
  4. Dashboard instability: Channels tab crashes on v0.6.8, degrading operational visibility (#5244)
  5. Security false positives: Legitimate commands like 2>/dev/null blocked by overzealous path scanner (#5518)

Positive Feedback

  • Telegram voice transcription support added (#5509 closed)
  • allowed_tools configuration option merged, improving operational flexibility (#5502)
  • Web UI improvements in progress for model selection clarity (#6218)

8. Backlog Watch

Long-Unanswered Important Issues Needing Maintainer Attention

Issue Title Age Risk Notes
#5146 Token consumption minimization via skill compilation ~32 days HIGH No PR yet; significant user interest
#5289 Bedrock provider sends API_KEY as x-api-key header ~26 days HIGH Status: in-progress
#5356 Canvas tool from channel server writes to separate CanvasStore ~25 days MEDIUM Frames never reach WebSocket clients
#5360 codex_cli tool passes unsupported -q flag ~25 days HIGH Fix PR #5361 exists
#5415 Context spillage from chat to schedule ~24 days S0 Security/data loss risk
#5470 Multiple issues when running safely ~23 days HIGH Blocked, needs repro

Stale PRs Requiring Review

PR Title Age Risk
#5972 fix: allow prompts inside [skill] TOML section ~9 days MEDIUM
#5978 feat(voice): speech capture buffer + STT dispatch ~9 days HIGH
#6009 feat(obs): enrich OTel tool spans ~8 days MEDIUM

Digest generated from Zeroclaw GitHub activity on 2026-04-30. Data sourced from github.com/zeroclaw-labs/zeroclaw.

PicoClawsipeed/picoclaw

Let me analyze the GitHub data for PicoClaw and generate a structured project digest for 2026-04-30.

Let me organize the data:

Releases:

  • Nightly build: v0.2.7-nightly.20260430.a36472b5

Issues (13 total, 12 open, 1 closed):

  1. #2171 - Enhancement: Refactor to use OpenAI Responses API (9 comments, open)

  2. #1042 - Bug: exec工具的guardCommand方法问题 (6 comments, open)

  3. #2548 - Bug: Multiple authentication credentials received (5 comments, open)

  4. #2625 - Enhancement: Provide compiled builds with WhatsApp support (2 comments, open)

  5. #2623 - Enhancement: .env file support (1 comment, open)

  6. #2621 - Bug: Session context lost after API timeout (1 comment, open)

  7. #2208 - Roadmap: Deprecate TUI version (1 comment, closed)

  8. #2671 - Enhancement: Support opencode provider (1 comment, open)

  9. #2706 - Bug: Deepseek v4 thinking model问题 (1 comment, open)

  10. #2720 - Bug: Singleton PID check doesn't verify process identity (0 comments, open, high priority)

  11. #2718 - Bug: DeepSeek 400 'unknown variant image_url' (0 comments, open)

  12. #2716 - Bug: SVG files fail to send on Telegram (0 comments, open)

  13. #2704 - Bug: DingTalk SDK panic (0 comments, open)

Now looking at the pull requests—there are 20 total with 15 open and 5 merged or closed. The open ones include a Slack webhook channel feature, a fix for tool feedback animation dismissal, and improvements to image handling in Feishu, among others.

Continuing through the open PRs, I'm seeing MCP header support for per-request dynamics, Bedrock streaming implementation, DeepSeek vision error detection, OpenAI-compatible embeddings, and a CI release workflow fix. There's also an important fix for the anthropic_messages provider around system content blocks with cache control, plus a multi-user group chat feature that's currently closed.

Looking at the rest of the queue, there's a web frontend copy button fix, Docker build restoration, Intel OpenVINO Model Server support, MQTT channel integration, reasoning content persistence for seahorse, and a channel identification refactor. The closed PRs include the multi-user group chat feature, another copy button fix, CLI support with TUI removal, and what appears to be a duplicate copy button fix.

Now I'm drafting a structured digest to organize all this information.

PicoClaw Project Digest — 2026-04-30


1. Today's Overview

PicoClaw shows high activity on 2026-04-30, with 13 issues and 20 PRs updated in the last 24 hours. A new nightly build (v0.2.7-nightly.20260430.a36472b5) was published. The community is actively contributing channel integrations (Slack webhook, MQTT), provider enhancements (Bedrock streaming, OpenAI embeddings, Intel OpenVINO), and bug fixes across channels and agents. Several high-priority bugs were reported today, including a singleton PID crash loop and a DingTalk SDK panic causing gateway crashes. The project is in a healthy but fast-moving state, with 15 PRs open and 5 merged/closed in the last day.


2. Releases

Nightly Build Published

  • Version: v0.2.7-nightly.20260430.a36472b5
  • Type: Automated nightly build (unstable)
  • Note: Users are advised to use with caution. Full changelog available at: sipeed/picoclaw@v0.2.7...main

3. Project Progress

Merged / Closed PRs (5)

# Title Domain Key Outcome
#2710 Add CLI support for custom OpenAI-compatible endpoints and remove TUI CLI / TUI Major CLI enhancement; TUI officially deprecated
#2714 Attribute history messages per sender for multi-user group chats Agent Closed in favor of #2715
#2711 Fix frontend copy button in HTTP (non-secure) context Web UX fix merged
#2709 Fix frontend copy button in HTTP (non-secure) context Web Duplicate fix (merged)
#2700 Restore make docker-build Docker Build tooling fixed

Notable Open PRs Advancing

# Title Domain Significance
#2719 Add slack_webhook output-only channel Channel New Slack integration via Incoming Webhooks
#2715 Attribute history messages per sender for multi-user group chats Agent Splits #2714 into 9 focused commits for review
#2705 Add MQTT channel support Channel New MQTT protocol integration
#2703 Add Intel OpenVINO Model Server support Provider Local LLM inference on Intel CPU/GPU/NPU
#2645 Implement StreamingProvider for Bedrock Provider Real-time token streaming via AWS ConverseStream API
#2624 Add OpenAI-compatible embeddings support Provider vLLM-style embedding endpoints supported
#2696 Support per-request dynamic headers from channel context (MCP) MCP Channels can forward HTTP headers to MCP servers
#2708 Fix Feishu image download with API fallback Channel Image extraction from post/rich-text messages
#2713 Dismiss tool feedback animation when turn ends Channel/Agent Fixes indefinite animation goroutine on Feishu
#2717 Add DeepSeek vision unsupported error detection Provider Detects unknown variant image_url errors
#2707 Persist reasoning_content in seahorse sqlite history Provider Fixes DeepSeek reasoning content storage
#2192 Fix anthropic_messages sending system as flat string with cache_control Provider Restores Anthropic prompt caching functionality
#2551 Standardize channel identification, decouple name from provider type Channel Enables multiple instances of same provider

4. Community Hot Topics

Most Active Issues (by comments)

# Title Comments Type Link
#2171 [Refactor

⚠️ 内容超过 GitHub Issue 上限,完整报告见提交的 Markdown 文件。

Metadata

Metadata

Assignees

No one assigned

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions