mirror of
https://github.com/dograh-hq/dograh.git
synced 2026-06-19 08:28:10 +02:00
test(mcp): guard instructions.py against tool drift
The MCP `instructions` hint is static and baked into the client prompt, while tool names, signatures, and error codes are discovered dynamically via tools/list. The two had drifted: instructions restated stale signatures and an error-code enum that omitted schema_validation and trigger_path_conflict. - Trim instructions.py to tool names + call order; stop restating signatures and error codes the dynamic surface already carries. - Document each tool's full error_code contract in the save_workflow and create_workflow docstrings (the descriptions shipped via tools/list). - Add test_mcp_instructions_drift.py: every tool named in the guide must be registered, and every error_code a tool returns must appear in its description. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
parent
5762095edf
commit
8484e4bfaf
4 changed files with 170 additions and 30 deletions
|
|
@ -12,10 +12,10 @@ Execution flow mirrors `save_workflow`:
|
|||
4. Persist via `db_client.create_workflow` — workflow row + v1
|
||||
published definition in a single transaction.
|
||||
|
||||
Error codes surfaced to the LLM match `save_workflow`. An additional
|
||||
`missing_name` error is returned when the source omits
|
||||
`new Workflow({ name: "..." })` — the name is required and there is no
|
||||
prior workflow to fall back to.
|
||||
Each failure path returns an `error_code` via `_error_result`. Those
|
||||
codes and their meanings are documented in the `create_workflow`
|
||||
docstring (the description shipped to the LLM via `tools/list`); keep the
|
||||
two in sync — `test_mcp_instructions_drift.py` enforces it.
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
|
@ -86,6 +86,22 @@ async def create_workflow(code: str) -> dict[str, Any]:
|
|||
On success the new workflow is published as version 1. Use
|
||||
`save_workflow(workflow_id, code)` for subsequent edits — those go to
|
||||
a draft.
|
||||
|
||||
On failure the result has `created: false`, a machine-readable
|
||||
`error_code`, and a human-readable `error` (with file:line:column
|
||||
where the problem is locatable). Resubmit the full corrected source —
|
||||
patches are not accepted. Possible `error_code` values:
|
||||
- `parse_error` — disallowed construct or malformed TypeScript.
|
||||
- `validation_error` — node data failed spec validation (unknown
|
||||
field, missing required, wrong type, option out of range).
|
||||
- `schema_validation` — wire-format (DTO) rejection; rare.
|
||||
- `graph_validation` — structural rule broken (e.g. no start node,
|
||||
unreachable node, edge to/from the wrong node type).
|
||||
- `missing_name` — `new Workflow({ name })` is absent or empty; the
|
||||
name is required and there is no prior workflow to fall back to.
|
||||
- `trigger_path_conflict` — a trigger node's path is already used by
|
||||
another workflow in this organization; rename it and resubmit.
|
||||
- `bridge_error` — internal/transient; retry once, then surface it.
|
||||
"""
|
||||
user = await authenticate_mcp_request()
|
||||
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue