Skip to content

Add consolidated manage_serving_endpoint tool (CRUD, logs, metrics, permissions)#413

Closed
jralfonsog wants to merge 4 commits intodatabricks-solutions:mainfrom
jralfonsog:feat/serving-crud
Closed

Add consolidated manage_serving_endpoint tool (CRUD, logs, metrics, permissions)#413
jralfonsog wants to merge 4 commits intodatabricks-solutions:mainfrom
jralfonsog:feat/serving-crud

Conversation

@jralfonsog
Copy link
Copy Markdown

@jralfonsog jralfonsog commented Apr 2, 2026

Summary

  • New consolidated MCP tool: manage_serving_endpoint(action=...) — 8 actions covering the full endpoint lifecycle
  • Core library: 8 new functions for serving operations (create, update, delete, logs, metrics, permissions)
  • Skill update: Updated databricks-model-serving with consolidated tool examples

Consolidation (replaces PRs #414, #415, #416)

Previously 8 separate tools across 4 PRs. Now 1 tool with 8 actions for token efficiency:

Action What it does Timeout
create Create endpoint with ServedEntityInput config 180s
update Update served entities, traffic, scaling 180s
delete Delete endpoint + manifest cleanup 180s
get_build_logs Retrieve build/deployment logs 180s
get_server_logs Retrieve runtime server logs 180s
export_metrics Export Prometheus metrics (parsed) 180s
get_permissions Get endpoint ACLs 180s
update_permissions Set endpoint ACLs 180s

Existing upstream tools (get_serving_endpoint_status, query_serving_endpoint, list_serving_endpoints) unchanged.

Includes

  • Identity tagging (get_default_tags, with_description_footer) on create
  • Manifest tracking (track_resource, remove_resource) on create/delete
  • Idempotent create (returns existing with created: false)
  • Prometheus metrics parsing into structured dicts

Test plan

  • 20 core lib unit tests — all passing
  • 15 MCP tool unit tests — all passing
  • Ruff lint + format pass
  • Integration test against aws-fe (pending)

This pull request was AI-assisted by Isaac.

…rmissions

Add 8 new functions to databricks_tools_core.serving: create, update,
delete endpoints; build/server logs; metrics export with Prometheus
parsing; get/update permissions. All functions follow existing patterns
with typed SDK exceptions, Google-style docstrings, and Dict returns.

Co-authored-by: Isaac
Add manage_serving_endpoint(action=...) with 8 actions: create, update,
delete, get_build_logs, get_server_logs, export_metrics, get_permissions,
update_permissions. Keeps upstream's 3 read-only tools unchanged. Includes
manifest tracking, identity tagging, and idempotent create.

Co-authored-by: Isaac
Add 20 core lib tests (CRUD, logs, metrics, permissions, Prometheus
parsing) and 15 MCP tool tests covering all 8 actions including
validation, idempotency, manifest tracking, and case insensitivity.

Co-authored-by: Isaac
Add manage_serving_endpoint action table and workflow examples for
create, update, debug (build/server logs), and permissions management.

Co-authored-by: Isaac
@jralfonsog jralfonsog changed the title Add Model Serving CRUD tools (create, update, delete endpoints) Add consolidated manage_serving_endpoint tool (CRUD, logs, metrics, permissions) Apr 6, 2026
@calreynolds
Copy link
Copy Markdown
Collaborator

Thank you for the contribution! While we don't think this content aligns with what we're looking for at this time — we're actively monitoring tool and skill bloat, and require use of our testing framework for contributions to be merged — we appreciate you opening these PRs! While I can't see your name on GitHub, if you're at Databricks feel free to reach out on Slack. 👍

@calreynolds calreynolds closed this Apr 8, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants