# MMDC documentation

> Documentation of the Markarian Multiwavelength Data Center (mmdc.am), its website and the astro-mmdc Python SDK

MMDC (https://mmdc.am), the Markarian Multiwavelength Data Center, builds multiwavelength,
multitemporal SEDs and light curves of blazars and of any sky position from radio to gamma rays,
runs on-demand Swift UVOT/XRT analysis, and fits SSC, EIC and hadronic blazar emission models.
It is used from the website or from Python with the astro-mmdc SDK (`pip install astro-mmdc`).

Rules for programmatic clients (scripts, pipelines, AI agents):

- Positions are ICRS right ascension and declination in degrees.
- Resolve a name first: `client.catalogs.autocomplete("Mrk 421")` returns entries with name, ra,
  dec, refcatalog. Over raw HTTP: POST https://mmdc.am/api/autocomplete/ with {"query": "Mrk 421"}
  answers {"results": [...]}. astropy's SkyCoord.from_name also works.
- Identify yourself. The SDK sends User-Agent astro-mmdc/<version>; `MMDC(app="your-project")`
  adds an X-MMDC-Client header. Raw HTTP must send X-MMDC-Client, otherwise the modeling POSTs
  answer HTTP 426.
- An SED request reuses any job within 2 arcsec; do not send `refresh` routinely.
- Respect Retry-After, keep at most 2 jobs in flight, and wait with the header `Prefer: wait=25`
  (the server maximum) instead of polling in a tight loop.
- Undated catalogue values carry MJD 50000 (`undated` true); they are not 1995 observations.
- Cite Sahakyan et al. 2024, AJ, 168, 289 (doi:10.3847/1538-3881/ad8231). MMDC data are
  licensed under CC BY 4.0 and require attribution.

Recipes (Python SDK, `client = MMDC(app="your-project")`):

- SED: `client.sed.get(ra=..., dec=..., name=...)`, then `.between(mjd_start, mjd_end)` and `.to_csv(path)`
- Light curve: `client.observations.cone_search(ra=..., dec=..., radius_arcsec=10, is_lightcurve=True)`
- Model curve: `client.modeling.infer(z=..., ebl=True, model_type="SSC", parameters={...})`
- Fit: `client.modeling.submit_batch(csv, z=..., ebl=True, model_type="SSC")`, then `client.modeling.wait_for_batch(id)`
- Swift: `client.madam.analyze(ra=..., dec=..., mjd_start=..., mjd_end=..., instrument="uvot")`

The full guide for scripts and agents, with raw HTTP examples and limits, is the first page below.


## Agents and scripts

- [Scripts and AI agents](https://docs.mmdc.am/agents/index.md)

## Python SDK

- [Getting started](https://docs.mmdc.am/getting-started/index.md)
- [Advanced](https://docs.mmdc.am/guides/advanced/index.md)
- [Blazar emission modeling](https://docs.mmdc.am/guides/modeling/index.md)
- [Observations](https://docs.mmdc.am/guides/observations/index.md)
- [Results & errors](https://docs.mmdc.am/guides/results-errors/index.md)
- [SED data](https://docs.mmdc.am/guides/sed/index.md)
- [Swift UVOT/XRT analysis](https://docs.mmdc.am/guides/swift/index.md)
- [API reference](https://docs.mmdc.am/api/index.md)

## Website

- [Fitting your SED](https://docs.mmdc.am/web/fitting/index.md)
- [Blazar SED modeling](https://docs.mmdc.am/web/modeling/index.md)
- [Getting SED data](https://docs.mmdc.am/web/sed-data/index.md)

## Citing

- [Citing](https://docs.mmdc.am/citing/index.md)