Studio Files

Files

Project → Files lists everything the project keeps in storage that is not a document: the files you upload and the output files your flow runs produce. It also lists the project's codeframes — documents, but where you would look for them — and opens them in the codeframe editor. This page covers uploading, what Studio reads from a data file you upload, downloading and deleting, the storage quota, reading an uploaded data file in a flow, codeframes, and what files cannot be used for.

---

The Files screen

Files   assets · run outputs                                        [ Upload ]

A Data file node in a flow reads an upload by its name. Upload here or from the node.
Each data file is read after upload: its rows and columns, and any columns that
look like personal data, show below.
Download links last 5 minutes. Don't paste them into the survey.

Codeframes  analysis/*.codeframe.json · coding open answers by hand and by rules
                                                          [ + New codeframe… ]
 Codeframe                      Codes      Themes   Updated
 why                            why        4        Sep 21, 2026        edit
   analysis/why.codeframe.json

Uploads and run outputs
 Name                         Size       Updated
 panel_wave2.csv              84.6 KB    Sep 28, 2026   [copy] [download] [delete]
   assets/panel_wave2.csv
   1,200 rows × 24 columns · possible missing codes in 3 columns
   ● Looks like personal data ⓘ
 report.html                  212.4 KB   Sep 20, 2026          [download] [delete]
   outputs/tables/report.html
 report_fig_1.png             31.0 KB    Sep 20, 2026          [download] [delete]
   outputs/tables/report_fig_1.png
 household_survey.xlsx        1.2 MB     Sep 12, 2026   [copy] [download] [delete]
   assets/household_survey.xlsx
   160 rows × 16 columns

The Codeframes section is described below. Under Uploads and run outputs:

Column	Shows
Name	the file name, with its stored path underneath (for example assets/panel_wave2.csv for an upload, outputs/tables/report_fig_1.png for a run output), and an icon for images (.png, .jpg, .jpeg, .gif, .svg, .webp), reports (.md, .html) and other files. Under an upload a Data file node can read, what Studio read from it, and a Looks like personal data mark when some of its columns look like personal data (see What Studio reads from a data file). On a phone, the size and date show under the name instead of in their own columns
Size	the stored size
Updated	the date the file's content last changed. For an upload, uploading new content under the same name changes it, uploading the same content again does not. For a run output, a run that writes the file with new content changes it, and a run that writes the same content again does not

The list is newest first by Updated, so an output that a run has just rewritten with new content moves to the top.

Each row has icon buttons:

Button	Tooltip	What it does
Copy (uploads only)	"Copy the path a flow reads it by: assets/name"	copies assets/<name>, the path a flow stores for the upload, to the clipboard ("Copied to clipboard"). A Data file node lists the uploads by name, so you need the path only in your own code
Download	"Download (the link lasts 5 minutes)"	downloads the file (below)
Delete	"Delete"	deletes the file after a confirmation (below)

With no files the section says No files yet — "Uploaded files (a data file for a flow to read, a sample) and the outputs your flows produce show up here." — with Upload a file.

Because the stored path is shown under each name, two files with the same name from different flows (outputs/tables/report_fig_1.png, outputs/brand/report_fig_1.png) can be told apart in the list.

Upload a file

1. Click Upload (or Upload a file on an empty screen).
2. In Upload file, drag a file onto the box ("Drag & drop a file here, or click to browse") or click it to choose one. The box then shows the name and size. Above the box the dialog says which files a Data file node reads: "A Data file reads CSV and other text tables (.csv, .tsv, .txt; any encoding; separated by commas, semicolons, tabs or vertical bars), Excel (.xlsx, .xlsm, .xls), SPSS (.sav), Stata (.dta) and Parquet files, up to 50 MB."
3. Before you upload, the dialog says the name Files will store the file under when that is not the name it has (see below), and anything else worth knowing about this file.
4. Click Upload. You see "Uploaded name", with the stored name. The file is stored as assets/<name>, and a Data file node lists it under that name.

You can also upload from a Data file node: Upload… under its File list opens the dialog as Upload a data file (or Upload a dictionary under Dictionary (JSON)). There it takes only what the field can use — a file of another kind, even one dragged onto the box, is refused and Upload stays off — and once the file is stored, the node chooses it. See Where files go.

Rules:

• Any file type is accepted here, one file per upload, up to 50 MB. A larger file is refused in the dialog before it is sent — "name is 52.0 MB; Files takes up to 50 MB. Save it as Parquet, or split it." for a data file — and an empty one with "empty file".
• A file a flow cannot read as data — a picture for a report, a PDF, a codebook — is uploaded and kept like any other, and the dialog says so quietly: "A flow can't read logo.png as data; it's kept in Files."
• A workbook over 20 MB gets a warning: "A workbook this large reads slowly: saved as CSV or Parquet it reads in a fraction of the time."
• Names are cleaned. Cyrillic letters (Russian and Ukrainian) are spelled in Latin and accents are dropped (café.csv → cafe.csv). No other letters are spelled in Latin: other Cyrillic letters (Serbian ђ, Kazakh қ, Belarusian ў), Greek letters, and letters such as ß, æ, ø and ł are left to the next step. Then every run of characters other than Latin letters (A–Z, a–z), digits, ., _ and - becomes _, leading and trailing dots and underscores are removed, and the name is cut to 128 characters, keeping its extension: Logo final (v2).png is stored as Logo_final_v2_.png, Волна 2.csv as Volna_2.csv, Straße.csv as Stra_e.csv. A name with nothing left in Latin letters, such as one in Chinese or Greek, is stored as upload_<8 characters>.<extension> — the same for the same name. A browser sends a double quote in a file name as %22, so Survey "A".csv is stored as Survey_22A_22.csv.
• The dialog says the stored name in one sentence that says what happened to this name, such as "It will be saved in Files as Volna_2.csv: Cyrillic letters are spelled in Latin and spaces become _.", "It will be saved in Files as Stra_e.csv: letters Studio can't spell in Latin become _." or "Studio can't spell this name in Latin letters, so it will be saved in Files as upload_431a963b.csv." A name with nothing to keep, such as ..., is refused ("Nothing of this name can be kept: …"), and so is a name with .. in it ("A flow can never read a file whose name has “..” in it. …"): no run would ever bring such a file in.
• The same name replaces the existing file. The dialog warns first: "Files already has name: this upload replaces it, and every flow that reads it reads the new file from its next run." Rename the file on your computer first if you want to keep both.
• A data file is read at once — see What Studio reads from a data file.
• Uploading and deleting need the member role or higher; any member can list and download.

Storage quota

Stored files count against a per-organization quota — the total of all files in all projects of the organization, including run outputs:

Plan	Storage
Free	250 MB
Plus	5 GB
Pro	50 GB
Corporate	unlimited

An upload that would exceed it is refused with "the Plan plan allows up to N MB of stored files; delete files or upgrade to add more". Replacing a file only counts the difference in size. The quota is checked when you upload; flow runs still store their outputs when you are over it.

Download a file

Download fetches the file through a temporary link that is valid for 5 minutes and always downloads (the browser does not open it in a tab). You see "Downloaded name".

Delete a file

Delete asks for confirmation:

Delete file — Delete name from this project's files. Permanent — there is no trash. Anything that reads this path (a flow's Data file node, a report figure) will fail on its next run.

Click Delete; you see "Deleted name". Deleted run outputs come back the next time their flow runs.

Run outputs

Every file a flow run produces — reports, charts, tables — is stored and listed here:

• a flow's outputs are kept under outputs/<flow>/<file> and replaced every time that flow runs, so Files always holds the latest run's version (earlier runs keep their own file lists in the run history); its Updated date moves only when the run changed its content;
• Run all stores each flow's report (the .md, its .html and the figures it shows) under outputs/<flow>/ as soon as that flow finishes, as a run of that flow alone would, and then its combined report, by default reports/report.md (set under Settings → Reports). When a flow fails, the reports of the flows that succeeded are still stored, and the combined report is written from them under the title "Combined report (incomplete)", opening with a section "Missing from this report" that names each flow left out and why (see The combined report);
• each run stores at most 50 files and 200 MB; files beyond that are skipped. When a run produces more, the report documents (.md, .html) are kept ahead of figures and other files, so a flow that draws many charts still stores its report.

Reports also appear on the Reports tab. See Analysis Flows and Reports.

---

What Studio reads from a data file

When you upload a file a Data file node can read (.csv, .tsv, .txt, .xlsx, .xlsm, .xls, .sav, .dta, .parquet), the engine reads it at once, in the same isolated sandbox flows run in, the way a Data file node reads it with its default options. Studio keeps what it found with the upload — never an answer, only what describes the file: its rows and columns, how it was read (format, encoding, delimiter, sheet, header rows), and for each column its name, label, type, scale, value labels, declared missing codes, the codes that look like missing codes, and whether it looks like personal data. A file of more than 50,000 rows is described from its first 50,000 (the row count then ends in +).

Under the file's name, the row then says:

• "rows rows × columns columns", and "· possible missing codes in n columns" when some columns hold codes that look like missing codes: negative codes such as -1 to -9, -97 to -99 or -997 to -999 in a column whose other values are not negative, or 97, 98, 99, 997 … far above a column's other codes. A column of real negative amounts, such as a balance, is not flagged. Until you declare them (the node's Missing codes, or Mark them as missing in these columns in its Columns panel), such codes count as answers: they show in frequencies and change a mean.
• Looks like personal data, a short warning mark, when some columns' names (in English or Russian) or values look like an email or IP address, a location, a name, a phone number, an address or a participant ID (Prolific, MTurk). Its ⓘ (point at it or click it; Esc closes it) names those columns, each with its kind — "email (e-mail)" — then, in a paragraph of its own, "Leave these out early, in one of two ways:", followed by a list: "in a flow: the Data file node can add a Select columns node without them" and "here in Files: delete the file and upload it without them". A screen reader reads the same words as the ⓘ button's description. It is a hint, never a drop: nothing is removed until you act (see Data files you upload).
• "Reading its columns…" while the file is being read. The screen asks again every few seconds until it can say what the file holds. A reading that has not answered in five minutes is dropped from the row; a Data file node that reads the file asks for it again.
• "A Data file node cannot read it as it is: …" with the reason, when the file cannot be read with the default options. The node's Reading options may still read it (see Data file).

Uploading new content under the same name reads the file again. A Data file node that reads the file with other options — a sheet, a delimiter, missing codes, a dictionary — has it read that way too, and keeps that reading beside the default one; it shows in the node, not on this screen.

Analyzing a dataset collected elsewhere

Upload a data file here and a flow can read it: a panel file, an earlier wave, a national survey, a dataset from another tool. Uploading a file never adds rows to the responses table.

1. Upload the file — here, or from the node in step 2. A CSV can come as Excel saves it with any regional settings: ; between the fields, a decimal comma, Windows-1251 or UTF-16 ("Unicode text"). If the file has a dictionary (<name>.dictionary.json), upload it too.
2. In the flow, add a Data file node and choose the file by name in its File list, such as panel_wave2.csv. Each upload in the list has its size and upload date under its name (for a data file Files has read, its rows × columns first), and a long name wraps; the note under the field gives the chosen file's rows × columns, size and upload date. If you uploaded its dictionary, Dictionary (JSON) offers it: click Use panel_wave2.dictionary.json.
3. Check the node's Columns panel: how the file was read, where its labels and scales come from, and its columns. If a title was read as the names or the wrong sheet was read, set the option it names under Reading options.
4. If the panel finds codes that look like missing codes, click Mark them as missing in these columns, and add a Missing values node below to turn them into blanks. If it shows Looks like personal data (its ⓘ names the columns), click Add a Select columns node without them beside it.
5. Add your analyses below. Their pickers list the file's columns under the heading "From panel_wave2.csv", each column's label and scale under its name; a list of more than ten has a filter box ("Filter variables…"). The questionnaire's variables are offered too only when the file is this survey's data.
6. Click Run to here on a node to see its result, or save and run the flow (or Run all).

The node, its options and the panel are described in Data file. Data collected in Qualtrics has a route of its own that keeps the survey's codes and labels: see Bringing in Qualtrics data.

A run, Run all, a scheduled run and Run to here / Preview all copy the uploads their flows name into the run before it starts — only those, not every file under Files. If one cannot be copied, the run's log says why:

Log line	Meaning
"note: assets/name is not among this project's Files"	nothing is uploaded under that name — check the spelling, or upload it
"note: assets/name is listed under Files but its content is gone"	the stored file is missing; upload it again
"note: uploads cannot be read (reason)"	storage is not reachable just now; try again later

When it copies an upload, the log says which version it read: "read assets/panel_wave2.csv (sha256 …)", the start of the content's hash, so two runs over a replaced upload can be told apart (see Which upload a run read).

The Data file node that reads a missing file then fails with the path it looked for. Replacing an upload (same name) takes effect on the next run; deleting it makes the next run fail.

A research bundle downloaded With the responses so far brings the uploads its flows name, at the same assets/<name> path, so the scripts find them there too; a bundle without data leaves them out, and its README says where to put them. See Reproducibility.

Tip. Another way in is to import the data into a project table with a connector — Supabase (Postgres) (Plus) or Your database (Postgres) (Pro) in the ← import direction — and read it with a Project table node. Imported columns arrive as text. See Importing a table.

Codeframes

The Codeframes section ("analysis/*.codeframe.json · coding open answers by hand and by rules") lists the project's codeframes: the coding schemes of its open questions, which a flow's Code open answers node applies. They are documents of the project, not uploads — each change is saved with a Save and shows in History — but this is where you would look for them.

Column	Shows
Codeframe	its name, with its path underneath (analysis/<name>.codeframe.json)
Codes	the open-text variable it codes
Themes	how many themes it has
Updated	when it was last saved

• edit on a row opens the codeframe in the codeframe editor; Back to Files returns here.
• New codeframe… starts one: it asks for the open-text variable and a name, and opens the editor. The new codeframe is listed here once you have saved it.
• With none: "No codeframe yet. A Code open answers node in a flow applies one; start it here or from the node."
• A viewer can open and read a codeframe; New codeframe… is disabled for them ("Your role in this project lets you read codeframes, not start one").
• There is no Delete for a codeframe; one that no node names does nothing.

See Coding Open Answers.

What Files is not for

Logos and question media

Current limitation. Files has no permanent public link. The only link it gives is the 5-minute download link, so a logo or image pasted from Files stops showing within minutes.

For the survey's Logo URL (Builder → Theme, or your organization's house style) and for a question's Media URL, use a stable public URL — for example an image on your institution's website or another public host. The hints say so: Logo URL — "a public https:// address of the image — a file under Files has no public link" (in the house style: "… a file under a project's Files has no public link"); Media URL (a question's Advanced section) — "image / video shown with the question — a public https:// address; a file under Files has no public link". See Theme and Branding.

See also

• Analysis Flows
• Reports
• Connectors
• Theme and Branding
• Limits and Quotas at a Glance

---

← Reproducibility: Code, Bundles and Citation · Studio contents · Connectors →