In the VLTI DataFlow the raw frames (FITS files) are classified and associated to be processed by the data reduction pipeline and to be distributed to the individual PIs. The associations are done using different tools :

• On Paranal: raw data are sent on the fly during the night on the pipeline workstation, they are classified and associated with the DO (Data organizer) and processed for quick-look purposes wihout requiring further processing or calibrations.
• In Garching: all frames including the day-time calibrations and/or calibrations taken on different nights are available when the science data of a night are processed. In this case, we can associate scientific data to calibration data which have been taken either before or after or even on a different night.

Each type of raw file needs to be classified in order to associate the raw files together with the appropriate pipeline recipe. The primary FITS header keywords (DPR keywords and some of the TPL keywords) are cross checked against a table. There are several steps in the associations (see also association map).

1. For the raw files with a DPR CATG=CALIB

• All the CALIB raw files with DPR TYPE = TRACK* and the same TPL START are associated with the files which carry DPR TYPE=PHOTOMETRY and the same TPL START (see Table 1). These files will be processed using the "reduce_fringe " pipeline recipe. The product will be a FITS file named PIPELFILE.pro.fits with PIPEFILE similar to the name of the first raw file in the association.
• the CALIB product is cross-checked against the VLTI calibrators catalogue. If the target is found in the catalogue, the pipeline recipe "transfer_function" is applied and an instantaneous intrumental transfer function is calculated and stored in the xxx_undispersed.pro.fits as well as in an ascii file.

1. For the raw files with a DPR CATG=SCIENCE

• the SCIENCE files are checked in the same way and all the files which have the same TPL START are associated. The product is then associated with a CALIB product (following the rule closest in time) to produce a calibrated Visibility using the pipeline recipe "calibrate_visibility"

The following tables shows the characteristics of the files associated together to be processed with the pipeline recipe "reduce fringe"