مهارات جديدة

Use when the user wants to build, initialize, validate, optimize, or refactor a model-powered assistant, internal tool, automation, evaluator, or workflow from a business scenario or common problem statement, including project-structure refactors or starter skeletons that may separate model setup, prompt config, and orchestration, even if the request also mentions a UI, app shell, or local model service such as Ollama, and it is still unclear whether the solution should stay a single request, add supporting capabilities, or become orchestration. The user does not need to mention Agently explicitly.

Use when the user wants Agently runtime extension capabilities: Action Runtime, built-in action packages, legacy tool compatibility, MCP access, Execution Environment lifecycle, FastAPIHelper or streaming API exposure, auto-function helpers, KeyWaiter, or optional agently-devtools observation, evaluation, and playground integration.

Use when the user is shaping Agently request-side behavior: model setup, settings files, prompt management, structured output, response reuse, streaming consumption, session memory, embeddings, knowledge-base indexing, retrieval, or retrieval-backed answers within one request family.

Use when the user wants to migrate an existing LangChain, LangGraph, LlamaIndex, CrewAI, or similar system into Agently, including choosing whether the source belongs to request/agent-side Agently behavior or TriggerFlow orchestration.

Use when the user needs Agently Dynamic Task, model-generated or app-submitted DAG planning, TaskDAG validation, DynamicTaskResolver handlers, or TaskDAGExecutor execution through Agently.create_dynamic_task. Dynamic Task is a first-class Agently API that uses TriggerFlow as an execution substrate.

Use when the user needs workflow orchestration such as branching, concurrency, approvals, waiting and resume, runtime stream, restart-safe execution, mixed sync/async function or module orchestration, event-driven fan-out, process-clarity refactors that make stages explicit, performance-oriented refactors that collapse split requests, or workflow definitions and chunk-level runtime metadata that must stay visible for debugging and visualization. The user does not need to say TriggerFlow explicitly.

Use when the user needs conversation continuity, memo, or restore-after-restart behavior for a request family, including session ids, chat history, request-side memory boundaries, and session-backed continuity.

Use when the user is shaping how one model request or request family should be instructed or templated, including prompt slots, input/instruct/info layering, mappings, recursive placeholder injection, prompt config, YAML or config-file-driven prompt behavior, and reusable prompt structure.

Use when the user wants to build, initialize, validate, optimize, or refactor a model-powered assistant, internal tool, automation, evaluator, or workflow from a business scenario or common problem statement, including project-structure refactors or starter skeletons that may separate model setup, prompt config, and orchestration, even if the request also mentions a UI, app shell, or local model service such as Ollama, and it is still unclear whether the solution should stay a single request, add supporting capabilities, or become orchestration. The user does not need to mention Agently explicitly.

Use when the user wants stable structured fields, required keys, value-level output validation, reliable machine-readable sections, or downstream-consumable output from one model request, including prompt-config-owned output contracts, `.output(...)`, tuple ensure flags, runtime `ensure_keys`, `.validate(...)`, and structured streaming.

Use when the request is already narrowed to wiring a model endpoint, env vars, settings-file-based model config, `${ENV.xxx}` placeholders, `auto_load_env=True`, or connectivity check for a model-powered feature, including local Ollama, Anthropic, dotenv-loaded DeepSeek or other OpenAI-compatible settings, plugin namespace placement, auth, request options, and minimal verification.

Use when the user wants to reuse one model result, read text/data/meta without re-requesting, or stream partial updates, including `get_response()`, async getters, `delta`, `instant`, and `specific`.