python-sdk
Operate the Argilla 2.x Python SDK for datasets, records, search, import/export, users, workspaces, webhooks, and SDK troubleshooting.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- python-sdk
- license
- Apache 2.0
- metadata
- { "disco-role": "operating" }
- description
- Operate the Argilla 2.x Python SDK for datasets, records, search, import/export, users, workspaces, webhooks, and SDK troubleshooting.
- disable-model-invocation
- true
02 资源文件 6
- 参考文档references/api-reference.md12.8 KB
- 参考文档references/data-formats.md7.8 KB
- 参考文档references/troubleshooting.md8.6 KB
- 参考文档references/workflows.md11.5 KB
- 脚本scripts/build_dataset_template.py8.6 KB
- 脚本scripts/webhook_listener_template.py7.5 KB
03 SKILL.md 正文
Argilla Python SDK
Use this sub-skill when a task asks you to use Argilla 2.x from Python: connect to an Argilla server, define dataset settings, log records, map incoming data to fields/questions/metadata/vectors, search/filter/export records, manage users/workspaces, or build a webhook listener.
Do not use this sub-skill for server deployment, Docker/Kubernetes, search-engine reindexing, OAuth/SSO service configuration, or legacy v1/Rubrix migration. Route those to the sibling server-ops or legacy-migration sub-skill.
First-choice references
- Read
references/api-reference.mdwhen you need verified Argilla 2.8.0dev0 SDK signatures, object relationships, collection accessors, or event names. - Read
references/workflows.mdwhen you need end-to-end recipes for connecting, creating datasets, logging data, searching/filtering, import/export, users/workspaces, or webhooks. - Read
references/data-formats.mdwhen you must format field/question values, record mappings, flatten/export columns, markdown/media, images, chat data, custom fields, metadata, vectors, or external ids. - Read
references/troubleshooting.mdwhen SDK calls fail due to credentials, API URL, default client, schema collisions, record mappings, images/chat/custom fields, vectors/search, Hub tokens, private Spaces, or webhooks.
Bundled helpers
- Run
scripts/build_dataset_template.py --helpwhen you need a safe offline generator that prints or writes a minimal Python dataset/settings/records template; the helper itself never contacts an Argilla server. - Run
scripts/webhook_listener_template.py --helpwhen you need a safe FastAPI webhook listener skeleton; it does not create, register, delete, or serve live webhooks unless the user explicitly passes live-server arguments.
Operating rules
- Treat the Python SDK as service-backed:
rg.Argilla(...),create,update,delete,records.log,to_hub,from_hub, and webhooks can contact a server or network service. - Prefer explicit clients. Create
client = rg.Argilla(api_url=..., api_key=...)before constructing SDK resources if environment defaults are not guaranteed, and passclient=clientto datasets and user/workspace/webhook resources where supported. - Keep dataset setting names unique across fields, questions, metadata properties, and vector fields. Use simple stable names, then map arbitrary source columns with
dataset.records.log(..., mapping=...). - Use
Record(id=...)or mapping targetidfor the record external id. The SDK serializes this asexternal_idinternally, but the public constructor/mapping target isid. - Use safe templates for webhooks. The
webhook_listenerdecorator registers webhooks when it is executed, so do not place it at module import time unless the user intentionally wants registration.
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).