matrx-scheduler
Server-side execution engine for the matrx scheduling spine (sch_* tables in
Supabase). Provides:
- a scanner loop that claims due tasks atomically,
- a runner that drives the host's agent_runner / tool_runner,
- an authoritative cron parser + next-due computation,
- and (since 0.3) an HTTP API surface (FastAPI) for task / trigger / run CRUD, manual fires, scanner status, and cron preview.
Any host application (aidream is the reference) wires it up via configure()
and either runs the scanner directly, mounts the HTTP routes, or both.
What it does (under the hood)
- Polls
sch_taskfor rows wherenext_due_at <= now()and the host surface is insurfaces[](or'any'). - Atomically claims a candidate by inserting a
sch_runwithclaim_tokenandclaim_expires_at(the lease). - Drives the matrx-ai agent runner (host-injected) against the claimed task,
or the host's tool_runner for
kind='tool'. - Writes back results (
status,result_summary,output_ref, etc.). - Recomputes the next fire time on recurring triggers; the DB cascade
updates
sch_task.next_due_atautomatically.
Capability-within, injection-without
matrx-scheduler doesn't import from a host app. It exposes a configure()
function the host calls at startup, passing in:
supabase_client— service-role client used by the SCANNER to read sch_task across users and update lease state.surface— the string this host identifies as insch_task.surfaces[](e.g.'server'for aidream,'desktop'for matrx-local).agent_runner— callable that runs an agent (kind='agent' tasks).tool_runner— optional callable for kind='tool' tasks. Hosts that don't claim tool tasks (e.g. aidream itself) leave this unset.user_supabase_factory— optional callable(user_jwt) -> AsyncClientused by the HTTP routes to build per-request user-scoped clients. If not provided, the package falls back to an env-based factory keyed onSUPABASE_MATRIX_URL+SUPABASE_MATRIX_PUBLISHABLE_KEY.get_app_context,emitter_factory,scan_interval_seconds,lease_seconds— see the_ext.configuredocstring.
Install with the host extra (pip install matrx-scheduler[host]) when
using it inside a full host app. Add the api extra
(pip install matrx-scheduler[api]) to mount the HTTP routes.
HTTP API
Available since 0.3. Optional — drop the import if you only want the scanner.
from fastapi import FastAPI
import matrx_scheduler
app = FastAPI()
# Host startup wiring (any host).
matrx_scheduler.configure(
supabase_client=service_role_client,
surface="server", # or "desktop", "extension", ...
agent_runner=my_agent_runner,
user_supabase_factory=my_user_client_factory, # optional
)
await matrx_scheduler.start_scanner()
# Mount the HTTP routes.
matrx_scheduler.api.include_routers(app, prefix="/scheduler")
Endpoints
All routes use matrx_connect.AppContext via Depends(context_dep). Every
write goes through a per-request Supabase client built from the caller's JWT;
RLS is the only line of authority on row ownership. The service-role client
(used by the scanner) is never accessible from these routes.
Cross-owner responses are expected authorization refusals. The defensive owner-mismatch diagnostic is logged at WARNING, not ERROR, so routine denied requests do not create production-failure telemetry families.
| Method | Path | Purpose |
|---|---|---|
| POST | /scheduler/tasks |
Create a task (optionally with agent_task + trigger) |
| GET | /scheduler/tasks |
List tasks (filter by kind, enabled) |
| GET | /scheduler/tasks/{id} |
Get task hydrated with agent_task, triggers, runs |
| PATCH | /scheduler/tasks/{id} |
Patch task fields (title, enabled, tags, etc.) |
| DELETE | /scheduler/tasks/{id} |
Soft-delete (set enabled=false) |
| POST | /scheduler/tasks/{id}/run-now |
Enqueue a manual run via sch_enqueue_manual_run |
| GET | /scheduler/triggers?task_id=... |
List triggers for a task |
| POST | /scheduler/triggers |
Create a trigger on an existing task |
| PATCH | /scheduler/triggers/{id} |
Patch a trigger (recomputes next_due_at if type/config changed) |
| DELETE | /scheduler/triggers/{id} |
Hard-delete a trigger |
| GET | /scheduler/runs?task_id=...&status=... |
List run history |
| GET | /scheduler/runs/{id} |
Get one run |
| POST | /scheduler/cron/validate |
Validate a cron expression + preview next N fires |
| POST | /scheduler/cron/preview-fires |
Preview next-fire times for any trigger config |
| POST | /scheduler/compute-next-due-at |
Compute the single next due_at for a trigger config |
| GET | /scheduler/status |
Scanner health (admin only) |
PATCH distinguishes an omitted nullable field from an explicit JSON null.
Sending {"description": null} or {"expires_at": null} clears that field;
omitting it leaves the stored value unchanged. This distinction is enforced
with Pydantic's model_fields_set, never by dropping every None value.
Co-existence with aidream's /scheduling/*
aidream had a small router at /scheduling/* (cron validation, run-now,
admin force-disable) before this surface existed. The package routes use
/scheduler/* (singular) so they don't collide. aidream can mount both
simultaneously, or migrate clients to the package routes opportunistically.
matrx-local mounts only the package routes; it has no /scheduling/* routes
to begin with.
See docs/SCHEDULING.md in the matrx-frontend repo for the full data-model
spec, trigger taxonomy, and lifecycle.
1 maintainer
Release files for matrx-scheduler 0.3.249
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| matrx_scheduler-0.3.249.tar.gz | 125.2 kB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| matrx_scheduler-0.3.249-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 215.3 kB
Release files / matrx_scheduler-0.3.249.tar.gz
| Download URL | matrx_scheduler-0.3.249.tar.gz |
|---|---|
| Size | 125.2 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
84097a280cf192c712a42413c73135490725fcca487e770bef3f865c2f81b175
|
|
BLAKE2b-256 checksum How to use checksums |
53f53f25d4406202f8b37478b184b6fda28d0d9581d059f9511881d665dff81f
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Sep 24, 2026.
Transparency logRelease files / matrx_scheduler-0.3.249-py3-none-any.whl
| Download URL | matrx_scheduler-0.3.249-py3-none-any.whl |
|---|---|
| Size | 90.2 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
b540e417ae41ed11882066d302d2c38b103bef2d96d2aa439bb72ff1db2ddd23
|
|
BLAKE2b-256 checksum How to use checksums |
c1bae497b9cfb6281dc621ab092efcbd319681a872020ab9b226445e96445d3b
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Sep 24, 2026.
Transparency log
