mirror of
https://github.com/hansjone/oclaw.git
synced 2026-10-09 01:50:44 +08:00
Default Weixin to official native runner, add /weixin/native/reply host endpoint, and harden gateway stop to clean stale port listeners. Document open-source installation flow for gateway + Personal WeChat. Made-with: Cursor
3 KiB
3 KiB
Oclaw Architecture Root
This repository is fully consolidated under oclaw/.
Quickstart (Open Source)
Prerequisites
- Python 3.11+
- Node.js 22+ (required by the official Weixin plugin)
1) Bootstrap venv (Windows)
powershell -ExecutionPolicy Bypass -File .\scripts\bootstrap_venv.ps1
2) Start gateway (background)
powershell -ExecutionPolicy Bypass -File .\scripts\start_gateway.ps1 -SkipInstall -Background
Open:
- Admin:
http://127.0.0.1:8787/admin - Chat:
http://127.0.0.1:8787/chat
If port 8787 is stuck/occupied, force-stop:
powershell -ExecutionPolicy Bypass -File .\scripts\stop_gateway.ps1 -Force
3) Weixin (Personal WeChat): install → login → start
powershell -ExecutionPolicy Bypass -File .\runtime\operations\scripts\weixin_install.ps1 -UseOpenclawCli
powershell -ExecutionPolicy Bypass -File .\runtime\operations\scripts\weixin_login.ps1
powershell -ExecutionPolicy Bypass -File .\runtime\operations\scripts\weixin_start.ps1
Full runbook (recommended): see docs/RUNBOOK.md → “开源快速安装(从零到跑起来)”.
Layers
runtime/: core execution loop, routing, skill runtime, hook runtimeinterfaces/: transport adapters (HTTP/WS)gateway/: method handlers and protocol bridgingapplication/: use-cases and orchestration servicesinfrastructure/: runtime-facing integrations/adaptersplatform/: shared platform capabilities (llm, persistence, config, files)tools/: tool registry, MCP adapters, public/system toolsskills/: installable skills and runtime manifests
Naming Rule
- Use
oclawconsistently in paths, symbols, and docs. - Avoid introducing legacy aliases or old naming variants.
Attachment Replay Config
- Attachment-related limits are configured in
oclaw.jsonunder:plugins.entries.memory-wiki.auto.attachments.tabular
- Replay limits:
image_result_replay_cap_chars(default4000, range600..30000)video_result_replay_cap_chars(default4000, range600..30000)- Used to cap historical
query_image_attachment/query_video_attachment(task=transcript)text replay in model context.
- Video transcript chunk defaults:
video_transcript_chunk_size(default1600)video_transcript_chunk_overlap(default200)
- Unified archive budget defaults (zip/tar/tgz/gz):
archive_max_depth(default2)archive_max_file_count(default200)archive_max_entry_bytes(default10485760)archive_max_total_uncompressed_bytes(default52428800)- Archive parse errors now expose stable
error_codevalues (for UI mapping and retries).
- Effective priority for replay-cap values:
- DB setting
AIA_IMAGE_TOOL_RESULT_REPLAY_CAP_CHARS - DB setting
AIA_VIDEO_TOOL_RESULT_REPLAY_CAP_CHARS - Environment variable
AIA_IMAGE_TOOL_RESULT_REPLAY_CAP_CHARS - Environment variable
AIA_VIDEO_TOOL_RESULT_REPLAY_CAP_CHARS oclaw.jsonvalue- Built-in default
- DB setting
See docs/ENVIRONMENT_VARIABLES.md for full runtime variable reference.