[Phase 1] Build unified normalization path for webhook + poller #7

New issue

Open

opened 2026-07-14 01:19:30 +00:00 by starsetbyte · 0 comments

starsetbyte commented

2026-07-14 01:19:30 +00:00

Owner

Context

The existing Forgejo issue poller and the new webhook path must feed the same idempotent normalization function. This prevents double-processing and ensures consistent event shape regardless of delivery mechanism.

What to build

Define a single normalize_forgejo_event() function that accepts:
- Source (webhook or poll)
- Raw Forgejo payload (for webhook) or issue/PR state (for poll)
- Delivery ID (webhook) or synthetic key (poll: repo:issue_number:action:updated_at)
The function produces the canonical normalized event with actor/origin fields.
Deduplication keys:
- Webhook: delivery_id from X-Forgejo-Delivery header
- Poll: repository:issue_number:action:updated_at hash
- Both produce an Aether correlation_id after normalization
The existing homelab-tagged issue polling loop becomes one ingestion rule:
- Issue with homelab or cross label -> translate to Aether offer/task-discovery event
- Does NOT independently assign work outside the Aether protocol
Both paths share the same SQLite deduplication table.

Acceptance

Single normalization function handles both webhook and poll inputs
Deduplication table shared across both paths
Poll-generated events produce valid correlation IDs
Existing poller behavior preserved (homelab/cross label detection)
No double-processing when webhook and poll both see the same event

## Context The existing Forgejo issue poller and the new webhook path must feed the same idempotent normalization function. This prevents double-processing and ensures consistent event shape regardless of delivery mechanism. ## What to build 1. Define a single `normalize_forgejo_event()` function that accepts: - Source (`webhook` or `poll`) - Raw Forgejo payload (for webhook) or issue/PR state (for poll) - Delivery ID (webhook) or synthetic key (poll: `repo:issue_number:action:updated_at`) 2. The function produces the canonical normalized event with actor/origin fields. 3. Deduplication keys: - Webhook: `delivery_id` from `X-Forgejo-Delivery` header - Poll: `repository:issue_number:action:updated_at` hash - Both produce an Aether `correlation_id` after normalization 4. The existing homelab-tagged issue polling loop becomes one ingestion rule: - Issue with `homelab` or `cross` label -> translate to Aether offer/task-discovery event - Does NOT independently assign work outside the Aether protocol 5. Both paths share the same SQLite deduplication table. ## Acceptance - [ ] Single normalization function handles both webhook and poll inputs - [ ] Deduplication table shared across both paths - [ ] Poll-generated events produce valid correlation IDs - [ ] Existing poller behavior preserved (homelab/cross label detection) - [ ] No double-processing when webhook and poll both see the same event

starsetbyte added the

phase-1

ingestion

labels

2026-07-14 01:19:30 +00:00

starsetbyte referenced this issue

2026-07-14 01:19:30 +00:00

[Phase 1] Retain existing poller as backstop — fold into new architecture #9