Skill: User Fastgpt

User FastGPT Workflow

Use this skill for FastGPT tasks aligned with vignettes/fastgpt-endpoint.qmd.

Required Workflow Order

1. Create kagi_connection().
2. Build query objects with query_fastgpt().
3. Prefer kagi_fetch() for project-folder workflows.
4. Use kagi_request() + kagi_request_parquet() for low-level control.

Allowed Function Set

• kagi_connection()
• query_fastgpt()
• kagi_fetch()
• kagi_request()
• kagi_request_parquet()

FastGPT-Specific Rules

• Keep prompts concise and task-specific.
• Use list query sets for repeatable prompt batches.
• Recommend error_mode = "write_dummy" for unattended runs.

References

Read and apply: - references/workflow.md - references/examples.md

References

Workflow

FastGPT Workflow (Aligned to Vignette)

Source of truth: vignettes/fastgpt-endpoint.qmd.

Sequence

1. Build connection.
2. Build query sets with query_fastgpt().
3. Preferred: execute end-to-end with kagi_fetch(project_folder = ...).
4. Low-level path: execute with kagi_request().
5. Use list queries for batch prompt runs.
6. Convert with kagi_request_parquet() when needed.

Error Strategy

• Use strict defaults for controlled runs.
• Use error_mode = "write_dummy" for unattended batches.

Constraints

• Keep prompts concise in examples.
• Keep behavior statements aligned with current docs/tests.

Examples

FastGPT Examples

q_fast <- query_fastgpt(
  query = "What is Python 3.11?",
  cache = TRUE,
  web_search = TRUE
)

kagi_request(
  connection = conn,
  query = q_fast[[1]],
  output = "fastgpt_output",
  overwrite = TRUE
)

kagi_request(
  connection = conn,
  query = q_fast_many,
  output = "fastgpt_batch_safe",
  overwrite = TRUE,
  workers = 2,
  error_mode = "write_dummy"
)