openwebui-importer

Import Grok, Claude, AI Studio and ChatGPT chats into open-webui.

This importer tool has two Python scripts: one for converting the model JSON files to openweb-ui format JSON, the second for creating a SQL script to import the JSON into the openweb-ui SQLite database.

The imported chats are given the tags imported-aistudio, imported-chatgpt, imported-claude and imported-grok.

Any private-use Unicode characters occasionally found in model exports are stripped from the message text during conversion.

There were problems exporting chats from Gemini, so it's not currently supported. DeepSeek and others could be added without much effort.

Quick start

python .\convert_chatgpt.py --userid="get-this-from-your-webui.db" .\chatgpt.json
python .\create_sql.py ./output/chatgpt --tags="imported-chatgpt" --output=chatgpt.sql
--
python .\convert_aistudio.py --userid="get-this-from-your-webui.db" .\aistudio_example.json
python .\create_sql.py ./output/aistudio --tags="imported-aistudio" --output=aistudio.sql

Quickstart Docker

docker run --rm -v $(pwd)/data:/data \
  ghcr.io/yetanotherchris/openwebui-importer:latest \
  python convert_chatgpt.py --userid="your-user-id" --output-dir=/data/output /data/chatgpt.json

docker run --rm -v $(pwd)/data:/data \
  ghcr.io/yetanotherchris/openwebui-importer:latest \
  python create_sql.py /data/output/chatgpt --tags="imported-chatgpt" --output=/data/chatgpt.sql

docker run --rm -v ${PWD}/data:/data `
  ghcr.io/yetanotherchris/openwebui-importer:latest `
  python convert_chatgpt.py --userid="your-user-id" --output-dir=/data/output /data/chatgpt.json

docker run --rm -v ${PWD}/data:/data `
  ghcr.io/yetanotherchris/openwebui-importer:latest `
  python create_sql.py /data/output/chatgpt --tags="imported-chatgpt" --output=/data/chatgpt.sql

Full example for GPT and Grok:

python .\convert_chatgpt.py --userid="example-9cef-4387-8ee4-b82eb2e1c637" .\chatgpt.json
python .\convert_aistudio.py --userid="example-9cef-4387-8ee4-b82eb2e1c637" .\aistudio_example.json
python .\convert_grok.py --userid="example-9cef-4387-8ee4-b82eb2e1c637" .\grok.json      
python .\create_sql.py ./output/chatgpt --tags="imported-chatgpt" --output=chatgpt.sql
python .\create_sql.py ./output/aistudio --tags="imported-aistudio" --output=aistudio.sql
python .\create_sql.py ./output/grok --tags="imported-grok" --output=grok.sql
# Now run the scripts inside DB Browser and hit save

Scripts

Install the required Python dependencies first:

pip install -r requirements.txt

convert_chatgpt.py

usage: convert_chatgpt.py [-h] --userid USERID [--output-dir OUTPUT_DIR] files [files ...]

Convert ChatGPT exports to open-webui JSON

convert_grok.py

usage: convert_grok.py [-h] --userid USERID [--output-dir OUTPUT_DIR] files [files ...]

Convert Grok exports to open-webui JSON

convert_claude.py

usage: convert_claude.py [-h] --userid USERID [--output-dir OUTPUT_DIR] files [files ...]

Convert Claude exports to open-webui JSON

convert_aistudio.py

usage: convert_aistudio.py [-h] --userid USERID [--output-dir OUTPUT_DIR] files [files ...]

Convert AI Studio exports to open-webui JSON

All converter scripts name the output files using the original conversation ID so running them again will produce the same filename for the same conversation. Converted files are saved in a subdirectory named after the model (for example output/grok or output/claude).

create_sql.py

usage: create_sql.py [-h] [--tags TAGS] [--output OUTPUT] files [files ...]

Create SQL inserts for open-webui chats. Existing chat records are deleted
before inserting so they are replaced if already present. Tags are inserted
with UPSERT statements, ensuring the default import tags (and any tags passed
via `--tags`) exist for each user.

positional arguments:
  files            Chat JSON files or directories

options:
  -h, --help       show this help message and exit
  --tags TAGS      Comma-separated tags for the meta field
  --output OUTPUT  Write SQL statements to this file

run_batch.py (Helper Script)

usage: run_batch.py [-h] --input-dir INPUT_DIR --type {aistudio,chatgpt,claude,grok} --user-id USER_ID [--output-dir OUTPUT_DIR] [--sql-output SQL_OUTPUT]

Batch process chat exports and generate SQL for Open WebUI.

options:
  -h, --help            show this help message and exit
  --input-dir INPUT_DIR
                        Directory containing the source chat files
  --type {aistudio,chatgpt,claude,grok}
                        Source chat format
  --user-id USER_ID     Open WebUI User ID to assign to these chats
  --output-dir OUTPUT_DIR
                        Directory for intermediate JSON files (default: output)
  --sql-output SQL_OUTPUT
                        Path to the final SQL file. If not specified, SQL generation is skipped.

Example:

python scripts/run_batch.py --input-dir ./my_chats --type aistudio --user-id "your-user-id" --sql-output aistudio_import.sql

This will convert all .json (and extensionless files for AI Studio) in ./my_chats and generate aistudio_import.sql.

Example workflow

1. Create an export from AI Studio (Gemini), Claude, ChatGPT or Grok.
2. Unzip the archive and locate the JSON file (for Grok this is prod-grok-backend.json).
3. Convert the export to open-webui JSON using the appropriate script:

   python ./convert_grok.py --userid="d95194d2-9cef-4387-8ee4-b82eb2e1c637" ./grok.json

   The converter writes JSON files to a subdirectory such as output/grok.
4. Generate SQL statements from the converted JSON files:

   python ./create_sql.py ./output --tags="imported-grok" --output=grok.sql

   The resulting SQL removes any existing chats with the same IDs before inserting new ones, while tags are inserted using UPSERTs so they are updated if they already exist. Any tags passed with --tags are also created for each user.
5. Make a copy of your webui.db database.
6. Execute the generated SQL using a tool such as DB Browser for SQLite. Ensure you save the database.