OpenWebUI becomes the interactive surface by talking to MOPAC like any
OpenAI provider: GET /v1/models lists the servable catalog (one model per
[models.classes] class, named mopac-<class>, routed through the same
tier table `once` uses; unknown model = 400 naming the valid ones) and
POST /v1/chat/completions runs ONE bounded stateless conductor turn over
the sent conversation history — no session storage, tools hard-off,
non-streaming (stream:true gets an explicit 400; OWUI tolerates
non-streaming providers). Bearer vkey auth compares SHA-256 digests in
constant time; missing/wrong keys get one byte-identical 401 body, and
the vkey never reaches logs or responses. temperature/max_tokens are
forwarded upstream; usage is summed across rounds and returned in the
reply. Upstream failures surface as a terse 502. Own port (:8090
default) so it coexists with the events receiver; dev.sh gets a serve
runner publishing 8090 on the LAN. Tests drive a scripted fake OpenAI
upstream through the real HTTP server: auth matrix, catalog + subset,
history assembly (client system message preserved, harness identity
prepended only when missing), multi-round usage accounting, refused
tool-call feedback, knob forwarding, 400/502 paths.
💘 Generated with Crush
Assisted-by: Crush:glm-5.2
README gains the self-host loop section (flow diagram, config reference
rows for [loop]/[redmine.status_map]/[keyproxy]/[gitea], CLI flags,
status table); REPORT.md records phase 3a including the die list for the
bash stack it replaces. dev.sh grows a `loop` runner (repo bind-mounted,
key env passthrough). `harness loop --once` now exits 2 when the intake
scan itself fails so cron setups alert loudly; daemon mode still logs
and retries.
💘 Generated with Crush
Assisted-by: Crush:glm-5.2
`harness loop` polls the /issues.json intake on an interval (default
120s, --once for cron-style single scans), runs one bounded turn per
new/updated issue — deduped by issue id + updated_on in an append-only
loop.jsonl — then writes the REPORT back as a Redmine journal note,
transitions status per [redmine.status_map] (In Progress -> Done by
name, resolved via /issue_statuses.json), and refreshes the dedup marker
to the post-writeback updated_on so its own notes never re-trigger it.
Turns are sequential (v0); failed turns are recorded, not retried, so a
down proxy cannot hot-loop the poll. The optional Gitea REPORT commit
([gitea] commit_reports, off) rides the same flow. No slot files, no
doorbell screens, no queue scripts — the bash middle layer is replaced,
not wrapped.
💘 Generated with Crush
Assisted-by: Crush:glm-5.2
The old path carried the reachableceo org prefix while the repo lives at
git.knownelement.com/ukrrs/MOPAC; the drift was cited as doc rot. go.mod
and every internal import updated; build/vet/test clean in the builder.
💘 Generated with Crush
Assisted-by: Crush:glm-5.2
`harness events` serves POST /hooks/{redmine,discourse,gitea} and
GET /healthz on stdlib net/http until SIGINT/SIGTERM (graceful
shutdown). Flow per delivery: verify (401, generic body) -> normalize
(400) -> append-only store with dedup (200 stored/duplicate + id +
action) -> hand stored actionable events to the conductor via the
Dispatcher interface. Conductor.DispatchEvent is the wiring point and
prints what it will do once phase 3 lands the real event-to-turn
dispatch. Body cap 1 MiB (413); audit log carries normalized fields +
digest only, never headers, secrets or payload. Server refuses to start
without at least one resolvable webhook secret.
`harness once` runs one bounded iteration (intake -> routing -> bounded turn
-> REPORT) then exits, so callers chain it without a daemon. The bounded
turn loops request -> tool calls -> results up to max_rounds; gate denials
feed back to the model and are counted instead of failing the turn.
--dry-run prints the plan (task, resolved model, tool bounds) and makes zero
LLM calls; --demo runs the [demo] issue ("tell me about yourself" through
LiteLLM -> GLM self-description as the REPORT), the MVP acceptance bar.
Distinct exit codes for config/usage, intake, and llm failures. Ships
harness.toml.example (scope query, tier map, allow-lists); real configs are
gitignored. Loop tests run end-to-end against a scripted fake OpenAI server:
demo turn, dry-run zero-call, tool round-trip, denial counting, round limit,
and error-class mapping.
💘 Generated with Crush
Assisted-by: Crush:glm-5.2
mrcharles