BINKY-WP-0006: close with three clean railiance mail cycles
Some checks failed
Work Records / validate (push) Has been cancelled

Host LLM_CONNECT_URL discovery uses ClusterIP when DNS is unreachable;
mail path fails hard if triage fails. Mark T06 done and workplan finished;
update residual docs (OperatingRhythm, cutover runbook, llm-rhythm).
This commit is contained in:
tegwick 2026-07-22 10:28:11 +02:00
parent 3d0815527e
commit 2fbbf9fc76
8 changed files with 75 additions and 43 deletions

View file

@ -48,7 +48,7 @@ state_hub_decision_id: "d47d288f-a76f-4e2f-b039-330028824397"
created_at: "2026-07-18" created_at: "2026-07-18"
consequence: "Workstation cron line removed 2026-07-21; railiance host timers consequence: "Workstation cron line removed 2026-07-21; railiance host timers
installed (scripts/railiance-rhythm/); three completion event types verified. installed (scripts/railiance-rhythm/); three completion event types verified.
Residual continuous intake → BINKY-WP-0006." Continuous intake residual closed by BINKY-WP-0006 (finished 2026-07-22)."
evidence: evidence:
- integrations/executor-cutover-runbook.md - integrations/executor-cutover-runbook.md
- integrations/harness-tenant-onboarding.md - integrations/harness-tenant-onboarding.md

View file

@ -58,11 +58,13 @@ all `enabled: true`). Execution on Railiance:
`23 8 * * 1-5 /home/worsch/binky-control/scripts/rhythm-session.sh # binky rhythm bridge` `23 8 * * 1-5 /home/worsch/binky-control/scripts/rhythm-session.sh # binky rhythm bridge`
and pause railiance timers. Script kept for one month as break-glass. and pause railiance timers. Script kept for one month as break-glass.
**Known residual (BINKY-WP-0006):** finish **llm-connect / OpenRouter** path **Railiance executor (BINKY-WP-0006 finished 2026-07-22):** mail path is
for mail triage + briefs on Railiance (structured LLM output + deterministic **mail-scan → mail-triage** via **llm-connect / OpenRouter** (structured JSON
apply). **Do not** install Claude Code or another coding agent on railiance01. + deterministic apply). Daily brief uses `brief-daily` the same way.
Mail-scan is already deterministic and proven there; see **Do not** install Claude Code or another coding agent on railiance01. Host
`integrations/railiance-llm-rhythm.md`. timers resolve `LLM_CONNECT_URL` via port-forward or Service ClusterIP (see
`integrations/railiance-llm-rhythm.md`). Optional residual: `brief-weekly`
subcommand for Friday review prep.
## Weekly review (founder, ~15 min) ## Weekly review (founder, ~15 min)

View file

@ -38,13 +38,14 @@ cd ~/binky-control && git pull
./scripts/railiance-rhythm/install-on-railiance.sh ./scripts/railiance-rhythm/install-on-railiance.sh
``` ```
## Residual (do not re-open T06) ## Residual after BINKY-WP-0006 (2026-07-22)
| Gap | Follow-up | | Gap | Status / follow-up |
| --- | --- | | --- | --- |
| Post-scan **triage** still assumes coding-agent session (Claude) | BINKY-WP-0006: **llm-connect → OpenRouter** + JSON apply (see `railiance-llm-rhythm.md`) | | Post-scan **triage** via llm-connect | **Done**`agent-harness mail-triage`; three clean railiance cycles 2026-07-22 |
| Daily/review unattended without host coding agent | Same pattern via llm-connect | | Daily brief without host coding agent | **Done**`brief-daily` |
| Host timers dual activity-core schedules | Simplify once llm path is the only executor | | Weekly review prep without host coding agent | **Open** — need `brief-weekly` (same pattern) |
| Host timers dual activity-core schedules | Simplify when in-cluster CronJob/poller owns cadence |
| **Out of scope / rejected:** install Claude Code on railiance01 | — | | **Out of scope / rejected:** install Claude Code on railiance01 | — |
## Rollback ## Rollback

View file

@ -1,7 +1,7 @@
# agent-harness tenant onboarding — binky-control (HARNESS-WP-0001-T07) # agent-harness tenant onboarding — binky-control (HARNESS-WP-0001-T07)
> Status: done 2026-07-18; cutover **BINKY-WP-0004-T06 completed 2026-07-21**. > Status: done 2026-07-18; cutover **BINKY-WP-0004-T06 completed 2026-07-21**;
> Residual continuous intake: BINKY-WP-0006. > llm-connect continuous intake **BINKY-WP-0006 finished 2026-07-22**.
## Instance manifest ## Instance manifest

View file

@ -1,6 +1,6 @@
# Railiance rhythm via llm-connect (not a host coding agent) # Railiance rhythm via llm-connect (not a host coding agent)
> Status: design note 2026-07-21 (BINKY-WP-0006). > Status: **live 2026-07-22** (BINKY-WP-0006 finished).
> Supersedes the residual “install Claude on railiance01” idea from the > Supersedes the residual “install Claude on railiance01” idea from the
> BINKY-WP-0004 cutover notes. > BINKY-WP-0004 cutover notes.
@ -36,16 +36,17 @@ timer / CronJob
- Output: `mailmeta/reports/email-channel-evidence-report-*.csv` - Output: `mailmeta/reports/email-channel-evidence-report-*.csv`
- Hub: `binky_mail_intake` with counts only - Hub: `binky_mail_intake` with counts only
### Phase 2 — triage (to build) ### Phase 2 — triage (**implemented**)
1. **Prompt** (metadata only): newest CSV rows + current mail-log tail + 1. **Prompt** (metadata only): newest CSV rows + current mail-log tail +
queue templates. queue templates.
2. **llm-connect** `POST` (same contract as activity-core `llm_client.py`): 2. **llm-connect** `POST /execute` (same contract as activity-core
- URL: `http://llm-connect.activity-core.svc.cluster.local:8080` `llm_client.py`):
- Provider: OpenRouter (already configured on railiance) - In-cluster: `http://llm-connect.activity-core.svc.cluster.local:8080`
- Model: cost-efficient; **prefer open-weights** (pick at T03; examples to - Host: Service ClusterIP (auto via `kubectl` in `run-common.sh`) or
evaluate: Llama/Qwen/Mistral class on OpenRouter — measure quality on port-forward `http://127.0.0.1:18080`
fixture CSV before locking). - Provider: OpenRouter; model currently `google/gemini-2.5-flash`
(open-weights retarget optional)
3. **Response**: strict JSON, e.g. 3. **Response**: strict JSON, e.g.
```json ```json
@ -59,10 +60,14 @@ timer / CronJob
``` ```
4. **Apply** (deterministic Python): append mail-log; add queue YAML only if 4. **Apply** (deterministic Python): append mail-log; add queue YAML only if
schema-valid; **suspicious** → log only; `git commit`; hub event. schema-valid; **suspicious** → log only; `git commit`; hub event
`binky_mail_triage`.
5. **Never** pass message bodies to the model if the scanner can omit them; 5. **Never** pass message bodies to the model if the scanner can omit them;
never execute model-suggested shell. never execute model-suggested shell.
**Dogfood (closure T06):** three clean unattended cycles on railiance01
2026-07-22 ~08:2608:27 UTC — each scan+triage `ok` with hub progress.
## Daily / weekly (same shape) ## Daily / weekly (same shape)
| Slot | Deterministic inputs | LLM output | Apply | | Slot | Deterministic inputs | LLM output | Apply |
@ -95,11 +100,12 @@ Hub event: `binky_daily_brief` with `detail.repo=binky-control`.
### Host reachability to llm-connect ### Host reachability to llm-connect
In-cluster DNS does not resolve on the bare host. Interim: In-cluster DNS does not resolve on the bare host. `run-common.sh` probes, in
order: port-forward `127.0.0.1:18080` → Service ClusterIP (via `kubectl`) →
cluster DNS. Manual override:
```bash ```bash
kubectl -n activity-core port-forward svc/llm-connect 18080:8080 & export LLM_CONNECT_URL=http://$(kubectl -n activity-core get svc llm-connect -o jsonpath='{.spec.clusterIP}'):8080
export LLM_CONNECT_URL=http://127.0.0.1:18080
export STATE_HUB_URL=http://127.0.0.1:18000 export STATE_HUB_URL=http://127.0.0.1:18000
agent-harness mail-triage --target-repo ~/binky-control agent-harness mail-triage --target-repo ~/binky-control
``` ```

View file

@ -30,14 +30,31 @@ require_repo() {
require_llm_connect() { require_llm_connect() {
# Server-side rhythm uses llm-connect HTTP (OpenRouter), never Claude CLI. # Server-side rhythm uses llm-connect HTTP (OpenRouter), never Claude CLI.
# Host cannot resolve in-cluster DNS; try reachable bases in order:
# 1) explicit LLM_CONNECT_URL 2) local port-forward 3) Service ClusterIP
# 4) cluster DNS (works only in-pod).
if [[ -z "${LLM_CONNECT_URL:-}" ]]; then if [[ -z "${LLM_CONNECT_URL:-}" ]]; then
# Host cannot resolve cluster DNS; default to local port-forward target. local candidates=()
# In-cluster Jobs should set LLM_CONNECT_URL to the cluster service. candidates+=("http://127.0.0.1:18080")
if curl -sf -m 1 "http://127.0.0.1:18080/health" >/dev/null 2>&1 \ if command -v kubectl >/dev/null 2>&1; then
|| curl -sf -m 1 "http://127.0.0.1:18080/" >/dev/null 2>&1; then local cip
export LLM_CONNECT_URL="http://127.0.0.1:18080" cip="$(kubectl -n activity-core get svc llm-connect \
else -o jsonpath='{.spec.clusterIP}' 2>/dev/null || true)"
export LLM_CONNECT_URL="http://llm-connect.activity-core.svc.cluster.local:8080" if [[ -n "${cip}" ]]; then
candidates+=("http://${cip}:8080")
fi
fi
candidates+=("http://llm-connect.activity-core.svc.cluster.local:8080")
local u
for u in "${candidates[@]}"; do
if curl -sf -m 2 "${u}/health" >/dev/null 2>&1; then
export LLM_CONNECT_URL="${u}"
break
fi
done
if [[ -z "${LLM_CONNECT_URL:-}" ]]; then
export LLM_CONNECT_URL="${candidates[-1]}"
log "WARN: no llm-connect health probe succeeded; using ${LLM_CONNECT_URL}"
fi fi
fi fi
log "LLM_CONNECT_URL=${LLM_CONNECT_URL}" log "LLM_CONNECT_URL=${LLM_CONNECT_URL}"

View file

@ -17,14 +17,14 @@ log "mail-scan rc=${scan_rc}"
if [[ "${scan_rc}" -ne 0 ]]; then if [[ "${scan_rc}" -ne 0 ]]; then
exit "${scan_rc}" exit "${scan_rc}"
fi fi
# Triage must use llm-connect (OpenRouter), not a host coding agent. # Triage uses llm-connect (OpenRouter), never a host coding agent.
# Until agent-harness mail-triage exists (BINKY-WP-0006), scan-only is OK. log "START mail-triage (llm-connect)"
if agent-harness mail-triage --help >/dev/null 2>&1; then agent-harness mail-triage \
log "START mail-triage (llm-connect)" --target-repo "${BINKY_REPO}" \
agent-harness mail-triage \ 2>&1 | tee -a "${LOG_DIR}/mail.log"
--target-repo "${BINKY_REPO}" \ triage_rc=${PIPESTATUS[0]}
2>&1 | tee -a "${LOG_DIR}/mail.log" || log "WARN: mail-triage failed" log "mail-triage rc=${triage_rc}"
else if [[ "${triage_rc}" -ne 0 ]]; then
log "SKIP mail-triage: command not available yet (BINKY-WP-0006); scan-only" exit "${triage_rc}"
fi fi
log "END mail-intake" log "END mail-intake"

View file

@ -4,7 +4,7 @@ type: workplan
title: "Railiance rhythm: llm-connect execution (no host coding agent)" title: "Railiance rhythm: llm-connect execution (no host coding agent)"
domain: infotech domain: infotech
repo: binky-control repo: binky-control
status: ready status: finished
owner: codex owner: codex
topic_slug: the-custodian topic_slug: the-custodian
created: "2026-07-21" created: "2026-07-21"
@ -155,10 +155,16 @@ grows, spawn BINKY-WP-0007 and mark this task done with handoff.
```task ```task
id: BINKY-WP-0006-T06 id: BINKY-WP-0006-T06
status: todo status: done
priority: low priority: low
state_hub_task_id: "3a93d7d4-f949-4a85-ad2e-a566ccc1f966" state_hub_task_id: "3a93d7d4-f949-4a85-ad2e-a566ccc1f966"
``` ```
Three clean unattended mail cycles on railiance (scan+triage); milestone; Three clean unattended mail cycles on railiance (scan+triage); milestone;
`statehub fix-consistency`. `statehub fix-consistency`.
**2026-07-22:** Three clean scan+triage cycles on railiance01 (UTC 08:2608:27),
each `ok` with hub events `binky_mail_intake` + `binky_mail_triage`. Host
`LLM_CONNECT_URL` discovery now prefers reachable ClusterIP (not broken
cluster DNS). Residual: `brief-weekly` still unimplemented (weekly timer will
fail until shipped); open-weights model retarget optional.