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

Define class-specific transport alert policy #103

Open
nsaspy wants to merge 10 commits from fix/61-alert-policy into fix/61-retry-routing-correlation
pull from: fix/61-alert-policy
merge into: nsaspy:fix/61-retry-routing-correlation
Conversation 0 Commits 10 Files changed 9 +637 -2
nsaspy commented 2026-09-20 06:13:36 +00:00
Owner
Copy link

Tracks the alert-policy acceptance of #61 as a separate bounded slice, stacked on #89.

What changed

  • add a deterministic transport alert-policy map from typed error class/scope to action, severity, recovery grace, rate class/window, and notification allowance;
  • downstream/client disconnects map to log_only / info / no notification allowance and never become provider alerts;
  • account/gateway failures map to immediate warning-level alert decisions with a 900-second account_gateway rate window and model attribution remains absent;
  • transient provider/router failures map to alert_if_unrecovered with the existing 10-second retry-correlation grace and a 60-second rate window;
  • stream protocol/integrity failures get a distinct protocol_integrity rate class with a 300-second window;
  • correlated retry recovery emits explicit suppress_recovered evidence linked to the original failed event;
  • add serialized append-only alert-decisions.jsonl projections containing safe policy metadata only—no request SHA and no router-attempt provider details;
  • wire the live completed transport-error/recovery correlation path through AlertPolicyActor with policy failures swallowed so capture/routing/recovery/forwarding cannot be blocked;
  • record the verified policy and boundaries in Org-roam and the project Prolog KB.

Downstream client disconnects intentionally bypass the completed-request correlation path established in #88; their deterministic policy is log_only and the typed client error remains in errors.jsonl. No external notification transport exists in this slice.

RED evidence

API contract head 3c3bf00eac16655c9307d961fee00ef21ef581c3: 127 tests ran. Exactly six new alert-policy tests errored because llm_log.alert_policy did not yet exist; every inherited #86-#89 contract stayed green. This is an executed missing-module RED, not a behavioral RED.

Behavioral wiring head bfae8fb3e864c71e21aa5e7f063b6e77c7eff07e: 128 tests ran. All six policy API tests passed and exactly one new live test failed because alert-decisions.jsonl was not produced. The existing suite remained green, isolating live wiring from policy semantics.

GREEN evidence

Live wiring head a9a8134cff36e4322f0bbf99d26d6fffa14c16a2 passed both exact-head workflows:

  • Analytics API run 71: full Python regression + adversarial analytics success;
  • Quota telemetry run 76: full Python regression + delivered package build success.

Final exact head 935fc1954f1b85cff1ea4579a8886e66c1df3add passed after Prolog/Org knowledge promotion:

  • Analytics API run 77: success;
  • Quota telemetry exact-head run 82: success, including delivered package build.

Boundary

This is a policy/decision stream, not a notification dispatcher. It does not send email, Discord, webhook, pager, desktop, or other external notifications and it does not own persistent notification-rate counters. A future dispatcher may enforce the emitted grace/rate policy. Model/structured-output anomaly detection remains separate #58 work.

No merge is performed here.

Tracks the alert-policy acceptance of #61 as a separate bounded slice, stacked on #89. ## What changed - add a deterministic transport alert-policy map from typed error class/scope to action, severity, recovery grace, rate class/window, and notification allowance; - downstream/client disconnects map to `log_only` / `info` / no notification allowance and never become provider alerts; - account/gateway failures map to immediate warning-level `alert` decisions with a 900-second `account_gateway` rate window and model attribution remains absent; - transient provider/router failures map to `alert_if_unrecovered` with the existing 10-second retry-correlation grace and a 60-second rate window; - stream protocol/integrity failures get a distinct `protocol_integrity` rate class with a 300-second window; - correlated retry recovery emits explicit `suppress_recovered` evidence linked to the original failed event; - add serialized append-only `alert-decisions.jsonl` projections containing safe policy metadata only—no request SHA and no router-attempt provider details; - wire the live completed transport-error/recovery correlation path through `AlertPolicyActor` with policy failures swallowed so capture/routing/recovery/forwarding cannot be blocked; - record the verified policy and boundaries in Org-roam and the project Prolog KB. Downstream client disconnects intentionally bypass the completed-request correlation path established in #88; their deterministic policy is `log_only` and the typed client error remains in `errors.jsonl`. No external notification transport exists in this slice. ## RED evidence API contract head `3c3bf00eac16655c9307d961fee00ef21ef581c3`: 127 tests ran. Exactly six new alert-policy tests errored because `llm_log.alert_policy` did not yet exist; every inherited #86-#89 contract stayed green. This is an executed missing-module RED, not a behavioral RED. Behavioral wiring head `bfae8fb3e864c71e21aa5e7f063b6e77c7eff07e`: 128 tests ran. All six policy API tests passed and exactly one new live test failed because `alert-decisions.jsonl` was not produced. The existing suite remained green, isolating live wiring from policy semantics. ## GREEN evidence Live wiring head `a9a8134cff36e4322f0bbf99d26d6fffa14c16a2` passed both exact-head workflows: - Analytics API run 71: full Python regression + adversarial analytics success; - Quota telemetry run 76: full Python regression + delivered package build success. Final exact head `935fc1954f1b85cff1ea4579a8886e66c1df3add` passed after Prolog/Org knowledge promotion: - Analytics API run 77: success; - Quota telemetry exact-head run 82: success, including delivered package build. ## Boundary This is a policy/decision stream, not a notification dispatcher. It does not send email, Discord, webhook, pager, desktop, or other external notifications and it does not own persistent notification-rate counters. A future dispatcher may enforce the emitted grace/rate policy. Model/structured-output anomaly detection remains separate #58 work. No merge is performed here.
docs: repair historical expert replay roam link
Some checks failed
Quota telemetry exact-head contract / quota-contracts (pull_request) Failing after 9s
Details
Analytics API / test (pull_request) Failing after 17s
Details
Analytics API / package (pull_request) Has been skipped
Details
935fc1954f
Some checks failed
Quota telemetry exact-head contract / quota-contracts (pull_request) Failing after 9s
Details
Analytics API / test (pull_request) Failing after 17s
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 fix/61-alert-policy:fix/61-alert-policy
git switch fix/61-alert-policy

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 fix/61-retry-routing-correlation
git merge --no-ff fix/61-alert-policy
git switch fix/61-alert-policy
git rebase fix/61-retry-routing-correlation
git switch fix/61-retry-routing-correlation
git merge --ff-only fix/61-alert-policy
git switch fix/61-alert-policy
git rebase fix/61-retry-routing-correlation
git switch fix/61-retry-routing-correlation
git merge --no-ff fix/61-alert-policy
git switch fix/61-retry-routing-correlation
git merge --squash fix/61-alert-policy
git switch fix/61-retry-routing-correlation
git merge --ff-only fix/61-alert-policy
git switch fix/61-retry-routing-correlation
git merge fix/61-alert-policy
git push origin fix/61-retry-routing-correlation
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!103
No description provided.