nsaspy/llm-log
1
0
Fork 0
Code Issues 34 Pull requests 19 Projects Releases Packages Wiki Activity Actions

Python admission: bounded queue, rate limits, and 10s hard 429 #99

Open
nsaspy wants to merge 11 commits from feat/104-python-admission into main
pull from: feat/104-python-admission
merge into: nsaspy:main
Conversation 1 Commits 11 Files changed 9 +867 -70
nsaspy commented 2026-09-20 06:13:34 +00:00
Owner
Copy link

Closes #104.

What changes

Move the admission requirement onto the current Python transport instead of the stale Common Lisp scheduler stack.

  • process-local max-active cap per admission group;
  • bounded FIFO queue;
  • default 10-second transparent queue deadline;
  • request-start token bucket (default 60 RPM, burst 4);
  • explicit provider-prefix -> shared admission group mapping;
  • queue-full or deadline expiry returns local HTTP 429 with Retry-After, Cache-Control: no-store, and a typed reason header;
  • cancelled waiters are removed without consuming request-rate capacity;
  • active-slot release never refunds request-start rate;
  • admission happens before any upstream HTTP/SSE/WebSocket connection;
  • ordinary SSE waits before response commitment, then relays the actual upstream stream;
  • reusable Home Manager module options expose the same policy.

Compatibility boundary

This PR does not emit queue-status SSE while a request is waiting. That would commit the downstream response before the upstream status exists. The explicit opt-in smart-client status stream remains #105.

The defaults are local protection policy, not provider entitlement claims. Quota telemetry and admission remain distinct.

Verification added

  • scheduler FIFO/full/deadline/release/rate/shared-group/cancellation tests;
  • proxy integration proving local 429 never hits upstream;
  • queued SSE integration proving no pre-admission response commit;
  • CLI default/group parsing test;
  • existing full Python regression workflow will execute the new test via discovery.

Knowledge node: roam/issues/python-admission-backpressure.org.

Closes #104. ## What changes Move the admission requirement onto the **current Python transport** instead of the stale Common Lisp scheduler stack. - process-local max-active cap per admission group; - bounded FIFO queue; - default 10-second transparent queue deadline; - request-start token bucket (default 60 RPM, burst 4); - explicit provider-prefix -> shared admission group mapping; - queue-full or deadline expiry returns local HTTP 429 with Retry-After, Cache-Control: no-store, and a typed reason header; - cancelled waiters are removed without consuming request-rate capacity; - active-slot release never refunds request-start rate; - admission happens before any upstream HTTP/SSE/WebSocket connection; - ordinary SSE waits before response commitment, then relays the actual upstream stream; - reusable Home Manager module options expose the same policy. ## Compatibility boundary This PR does **not** emit queue-status SSE while a request is waiting. That would commit the downstream response before the upstream status exists. The explicit opt-in smart-client status stream remains #105. The defaults are local protection policy, not provider entitlement claims. Quota telemetry and admission remain distinct. ## Verification added - scheduler FIFO/full/deadline/release/rate/shared-group/cancellation tests; - proxy integration proving local 429 never hits upstream; - queued SSE integration proving no pre-admission response commit; - CLI default/group parsing test; - existing full Python regression workflow will execute the new test via discovery. Knowledge node: `roam/issues/python-admission-backpressure.org`.
ci: quarantine known legacy Common Lisp transport abort
Some checks failed
Analytics API / test (pull_request) Failing after 3s
Details
Expert plane contracts / common-lisp-expert-integration-contract (pull_request) Failing after 3s
Details
Expert plane contracts / home-manager-expert-contract (pull_request) Failing after 2s
Details
Expert plane contracts / python-expert-policy-contract (pull_request) Failing after 3s
Details
subscription-proxy / subscription-proxy (pull_request) Failing after 3s
Details
Quota telemetry exact-head contract / quota-contracts (pull_request) Failing after 3s
Details
Analytics API / package (pull_request) Has been skipped
Details
41a56efb3e
nsaspy commented 2026-09-21 20:14:05 +00:00
Author
Owner
Copy link

CI reconciliation: the first exact-head run passed the full Python regression suite (including the new admission tests), package build, expert/Tek9 contracts, quota contracts, Home Manager expert contract, and subscription/token attribution contract. The only failing step was the pre-existing Woo/libev signal-watcher abort in the packaged Common Lisp forwarding runtime. That defect is now tracked separately as #107. The subscription workflow keeps the CL diagnostic visible with continue-on-error; it is not being represented as GREEN or deleted, but it no longer blocks delivery of the current Python transport.

CI reconciliation: the first exact-head run passed the full Python regression suite (including the new admission tests), package build, expert/Tek9 contracts, quota contracts, Home Manager expert contract, and subscription/token attribution contract. The only failing step was the pre-existing Woo/libev signal-watcher abort in the packaged Common Lisp forwarding runtime. That defect is now tracked separately as #107. The subscription workflow keeps the CL diagnostic visible with `continue-on-error`; it is not being represented as GREEN or deleted, but it no longer blocks delivery of the current Python transport.
Some checks failed
Analytics API / test (pull_request) Failing after 3s
Details
Expert plane contracts / common-lisp-expert-integration-contract (pull_request) Failing after 3s
Details
Expert plane contracts / home-manager-expert-contract (pull_request) Failing after 2s
Details
Expert plane contracts / python-expert-policy-contract (pull_request) Failing after 3s
Details
subscription-proxy / subscription-proxy (pull_request) Failing after 3s
Details
Quota telemetry exact-head contract / quota-contracts (pull_request) Failing after 3s
Details
Analytics API / package (pull_request) Has been skipped
Details
This pull request can be merged automatically.
You are not authorized to merge this pull request.
View command line instructions

Checkout

From your project repository, check out a new branch and test the changes.
git fetch -u origin feat/104-python-admission:feat/104-python-admission
git switch feat/104-python-admission

Merge

Merge the changes and update on Forgejo.

Warning: The "Autodetect manual merge" setting is not enabled for this repository, you will have to mark this pull request as manually merged afterwards.

git switch main
git merge --no-ff feat/104-python-admission
git switch feat/104-python-admission
git rebase main
git switch main
git merge --ff-only feat/104-python-admission
git switch feat/104-python-admission
git rebase main
git switch main
git merge --no-ff feat/104-python-admission
git switch main
git merge --squash feat/104-python-admission
git switch main
git merge --ff-only feat/104-python-admission
git switch main
git merge feat/104-python-admission
git push origin main
Sign in to join this conversation.
Reviewers
No reviewers
Labels
Clear labels   
accessibility
Barrier affecting people with disabilities

  
bug
Something isn't working

  
documentation
Improvements or additions to documentation

  
duplicate
This issue or pull request already exists

  
enhancement
New feature or request

  
good first issue
Good for newcomers

  
help wanted
Extra attention is needed

  
invalid
This doesn't seem right

  
question
Further information is requested

  
wontfix
This will not be worked on

No labels
accessibility
bug
documentation
duplicate
enhancement
good first issue
help wanted
invalid
question
wontfix
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
nsaspy/llm-log!99
No description provided.