rosa
Route ROSA LangChain agent construction, ROS 1 and ROS 2 operations, custom tool integration, and safe troubleshooting for jpl-rosa.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- rosa
- license
- Apache 2.0
- metadata
- { "disco-role": "operating" }
- description
- Route ROSA LangChain agent construction, ROS 1 and ROS 2 operations, custom tool integration, and safe troubleshooting for jpl-rosa.
- disable-model-invocation
- true
02 资源文件 5
- 参考文档references/repo-provenance.md1.8 KB
- 参考文档references/repo-routing-metadata.json399 B
- 参考文档references/runtime-prerequisites.md2.6 KB
- 参考文档references/troubleshooting.md4.9 KB
- 脚本scripts/check_environment.py3.7 KB
03 SKILL.md 正文
ROSA repository skill
Use this skill when a task names ROSA, jpl-rosa, from rosa import ROSA,
LangChain tool-calling agents for ROS, ros_version, ROSA tools, or the ROSA
TurtleSim-style agent pattern. It is a self-contained operating guide for the
public package; it does not require the original repository checkout.
Route the request
- Construct, configure, invoke, stream, or debug the agent: read agent-core.
- ROS 1 / Noetic graph, topics, nodes, services, parameters, packages, logs, launch, or kill: read ros1-operations.
- ROS 2 / Humble, Iron, or Jazzy CLI-backed graph, topics, services, parameters, doctor, or logs: read ros2-operations.
- Custom LangChain tools or packages, blacklists, robot prompts, math, logging, or safe adaptation: read tool-customization.
Start with the route matching the user's primary action, then follow its cross-links instead of loading every reference. Read runtime prerequisites and troubleshooting when environment boundaries or failures are involved. The spaces in the first link are intentional Markdown-label syntax; the target remains inside this skill tree.
Package boundary and minimal check
The public distribution is jpl-rosa and the import package is rosa.
Install it in Python >=3.9,<4:
python -m pip install jpl-rosa
python -c "from rosa import ROSA, RobotSystemPrompts, ChatModel; print('rosa import ok')"
The base package includes LangChain OpenAI integration. Install only the
provider extra needed by the selected workflow, for example
jpl-rosa[anthropic] or jpl-rosa[ollama]. Package installation does not
install ROS 1/ROS 2 middleware, a ROS master/daemon, a DDS graph, an LLM
service, credentials, Docker, or a display server. Use the safe
environment checker before constructing a live
agent; it never contacts a model or ROS system unless an explicit ROS check is
requested.
Safe operating rules
- Select
ros_version=1or2from the user's actual middleware; never mix the ROS 1 and ROS 2 tool catalogs. - For action requests, discover current nodes and topics first, then inspect the exact service/topic/parameter/package before acting.
- Execute one tool call at a time. Do not batch movement, drawing, service, launch, kill, or parameter operations.
- Treat service calls, parameter writes, launches, and node kills as mutations; state the target and expected effect before execution.
- Use the package's calculation tools for angles, distances, coordinates, and geometry rather than doing robotics arithmetic manually.
- Keep API keys and private robot details out of prompts, logs, source files, and generated reports.
Provenance and limits
Read repo-provenance.md before deciding whether this skill matches a changed ROSA checkout. Read the shared troubleshooting reference for install/import, provider, middleware, input-validation, and demo limitations. ROSA's Docker/X11 TurtleSim demo is intentionally documented as a bounded optional integration; this skill does not launch it automatically.
04 同仓库更多 Skills
passnet-triton-opt
Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block/grid/warp tuning, autotune policy, launch overhead, and per-op numeric recipes to pass the dtype baseline tolerances. Use when a pass matches and is correct but speedup below expected, or when correctness fails by small numeric margins.
passnet-solve
END-TO-END playbook for solving one PassNet sample: analyze the computation graphs, decide the optimization strategy (what to fuse, what to replace, what to leave alone), drive the iteration loop, and maximize the sample score. This is the ENTRY skill — invoke it first for any PassNet optimization task; it tells you when to use passnet-pattern-fusion, passnet-triton-opt and passnet-feedback.
passnet-skill
PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under ./pass_dir/, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.
passnet-pattern-fusion
Author PassNet pass files: write patterns that actually MATCH the FX graph, pick fusion regions (which ops to absorb into one Triton kernel), and structure multi-pass submissions with the mandatory shared-dispatch architecture. Use when creating or fixing pattern()/replacement_args()/replacement_func() files, when a pass "failed to match", or when deciding how to fuse multiple kernels/ops.
passnet-orchestrate
Round/batch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep/revert decisions. For solving one concrete sample end to end, use passnet-solve as the entry skill; consult this skill only when you need broader planning or a specific decision gate.
passnet-feedback
Fast, deterministic feedback for PassNet work: pre-flight pattern/match verification WITHOUT burning a GPU evaluation, per-node bottleneck analysis of a sample's graphs, and eval-log parsing into per-variant status + estimated score + failure classification. Use BEFORE every GPU evaluation (check_pattern), at the START of a sample (analyze_graph), and AFTER every evaluation (parse_eval_log).