workforce-workflows
Guides OWL Workforce and CAMEL multi-agent workflows, provider selection, model configuration, worker/tool composition, and safe task execution planning.
- License 许可
- Apache-2.0
- 最近更新
- 2026-10-08
- Stars
- ★ 11
- Forks
- 3
01 前言 Front Matter
- name
- workforce-workflows
- license
- Apache 2.0
- metadata
- { "disco-role": "operating" }
- description
- Guides OWL Workforce and CAMEL multi-agent workflows, provider selection, model configuration, worker/tool composition, and safe task execution planning.
- disable-model-invocation
- true
02 资源文件 4
- 参考文档references/provider-recipes.md3.4 KB
- 参考文档references/role-playing-api.md2.9 KB
- 参考文档references/toolkit-composition.md2.9 KB
- 参考文档references/troubleshooting.md2.5 KB
03 SKILL.md 正文
OWL Workforce Workflows
Use this route when the task is to configure or run OWL's hierarchical CAMEL Workforce, select a remote model provider, assemble specialist workers, or understand the custom role-playing loop. It covers the provider-backed runtime; it does not claim that a model call or browser session is available.
Start here
- Install OWL in an isolated Python 3.10–3.12 environment and run the root environment check before constructing a model.
- Choose one provider in provider-recipes.md. Keep provider credentials separate from optional search, crawling, or document-service credentials.
- Validate a non-secret environment file with
validate_provider_config.py before
calling CAMEL's
ModelFactory. - Read toolkit-composition.md when the task needs browser, search, code, spreadsheet, file, image, or document tools.
- Use scaffold_workforce.py when a safe, non-executing configuration skeleton is more useful than copying a credentialed example.
- Build the Workforce and pass a
camel.tasks.Taskcontaining the complete user request. Use role-playing-api.md only when the task needs OWL's custom two-agent loop or GAIA mode.
Canonical workflow
- Create provider-specific model backends with
camel.models.ModelFactory. - Construct specialist models for web, document, reasoning, image, browsing, and planning roles as needed. A model that cannot call tools is not a valid substitute for a tool-intensive worker.
- Create toolkit instances, wrap individual methods with
FunctionToolwhen required, and add toolkit tools to the appropriateChatAgent. - Add worker agents to a
camel.societies.Workforcewith descriptions that make their boundaries discoverable. Add a task and callworkforce.process_task. - Pass a user question as the first command-line argument to a provider example
rather than editing a credential into code. For a reusable starting point,
generate a non-executing skeleton with
scaffold_workforce.py. - For local files, route document-specific behavior to document-processing. For the Gradio front end or Docker, route to web-ui-and-deployment.
Safety and limits
Remote provider calls, search, Firecrawl, browser automation, and code
execution can incur cost or perform external side effects. Confirm credentials,
network access, browser policy, and output paths before execution. Keep
CodeExecutionToolkit in its intended sandbox and do not expose unrestricted
file or terminal tools to untrusted tasks. A CPU import check proves only that
OWL and CAMEL import; it does not prove provider tool calling or multimodal
quality.
Read troubleshooting.md when import errors, missing keys, endpoint failures, stale example names, or tool capability mismatches appear. The source examples are evidence for this skill, not files a future agent must reopen or execute from the original checkout.
04 同仓库更多 Skills
obsidian-clipper-template-creator
Guide for creating templates for the Obsidian Web Clipper. Use when you want to create a new clipping template, understand available variables, or format clipped content.
obsidian-clipper-template-creator
Guide for creating templates for the Obsidian Web Clipper. Use when you want to create a new clipping template, understand available variables, or format clipped content.
python-api
Programmatic Object-Detection-Metrics API guidance for BoundingBox, BoundingBoxes, Evaluator, enums, coordinate conversion, AP/mAP semantics, and source-style import diagnostics.
file-evaluation
Folder-based PASCAL VOC AP/mAP evaluation from ground-truth and detection text files.
object-detection-metrics
Guides legacy Object-Detection-Metrics PASCAL VOC AP/mAP evaluation from detection text folders and direct Python API classes.
nx-workspace-patterns
Configure and optimize Nx monorepo workspaces. Use when setting up Nx, configuring project boundaries, optimizing build caching, or implementing affected commands.