> ## Documentation Index > Fetch the complete documentation index at: https://docs.hotglue.com/llms.txt > Use this file to discover all available pages before exploring further. # Running Targets > Learn how to run targets using the Singer SDK # Background A target is an integration that reads data and writes it to some *target* system. For example, hotglue supports targets for: * Uploading files to storage systems S3, Azure Blob Storage, or SFTP servers. * Creating and Updating invoices in ERPs like Netsuite, Sage Intacct, and Dynamics BC * Upserting records into databases like Postgres, Snowflake, and BigQuery In most cases, targets expect their input in the form of a [**Singer**](https://www.singer.io/) file, a variation of jsonl for structured streaming data. # Setting up a local target environment ## Virtual Environments Dependencies can differ from target to target, so it is best to use a virtual environment to isolate the target's Python dependencies. You can create a virtual environment named `.venv` in your target workspace with: ```bash theme={null} python -m venv .venv ``` You can then enter the virtual environment with: ```bash theme={null} source .venv/bin/activate ``` Finally, you can install the tap's dependencies depending on how they are specified: | Dependency File | Command | | :----------------: | :-------------------------------: | | `requirements.txt` | `pip install -r requirements.txt` | | `setup.py` | `pip install -e .` | | `pyproject.toml` | `pip install -e .` | ## Config Targets require a valid config, which will include fields like API keys, OAuth credentials, and configuration flags. For example: ```json theme={null} { "client_id": "...", "client_secret": "...", "refresh_token": "...", "access_token": "...", "start_date": "2025-01-25T00:00:00Z" } ``` Most target will include information about the required config values in their README's. > 💡 To keep your workspace clean, create a `.secrets` folder to store your `config.json`. Run all tap commands out of this folder to make sure your catalogs and output data stay separate from relevant code. ## Using the Hotglue Access Token Endpoint For OAuth connectors, you can have the target use the [access token endpoint](https://docs.hotglue.com/api-reference/access-tokens/retrieve-access-token) instead of refreshing the token locally by: **1. Adding this flag to your `config.json`:** ```json theme={null} { "_refresh_token_via_hg_api": true, ... } ``` **2. Setting the following environment variables:** | Variable | Description | | :-------: | :-------------------------------- | | `TENANT` | Your tenant ID | | `API_KEY` | Your Hotglue API key | | `FLOW` | The flow ID | | `ENV_ID` | The environment ID | | `TARGET` | The target ID (e.g. `salesforce`) | ```bash theme={null} export TENANT="" export API_KEY="" export FLOW="" export ENV_ID="" export TARGET="" ``` ## Data If you want to try to reproduce the behavior of a particular [hotglue job](/key-concepts/jobs/sync-jobs), you can download the job folder on your hotglue job page: The etl-output folder will contain the raw input file(s) used by the target. In most cases, you'll see a singular `data.singer` file with all your data: ```jsonl theme={null} {"type":"SCHEMA","stream":"Invoices","schema":{"type":["object","null"],"properties":{"invoiceId":{"type":["string","null"]},"date":{"type":["string","null"]},"total":{"type":["number","null"]}}},"key_properties":[]} {"type":"RECORD","stream":"Invoices","record":{"invoiceId":"INV-001","date":"2025-02-11T00:00:00.000Z","total":112.40}} {"type":"SCHEMA","stream":"InvoicePayments","schema":{"type":["object","null"],"properties":{"paymentId":{"type":["string","null"]},"amount":{"type":["number","null"]}}},"key_properties":[]} {"type":"RECORD","stream":"InvoicePayments","record":{"paymentId":"PAY-001","amount":85.50}} ``` Place your `data.singer` somewhere accessible by your target. We recommend placing it in a git-ignored `.secrets` with your `config.json` to avoid accidently committing it. # Running a target You can now pipe your data into the target. If your target is named `target-salesforce`, you would run: ``` target-salesforce < data.singer --config config.json > target-state.json ``` This will process your singer file line by line and (if supported by the target) write the export results to a file named `target-state.json`. If you attempted to send two contacts to Salesforce, your target state might look something like this: ```json theme={null} { "bookmarks": { "Contacts": [ { "hash": "372f2b3f76de48b1823db1ed2333042ef312ce2ba2b50165467ae64a78c05a71", "success": true, "id": 900, "externalId": "903093" } { "hash": "97a8sdhge48b1823db1ed2333042ef312ce2ba2b5016sd5467ae64a78c05a71", "externalId": "903094", "error": "Email 'Bi@ll@hotg.lue.com is invalid'" } ] }, "summary": { "Contacts": { "success": 0, "fail": 1, "existing": 0, "updated": 1 } } } ``` Here `id` is the id in Salesforce and `externalId` (if passed on the record) is the id in the source system. If you prefer to work with [VSCode's Debugger](https://code.visualstudio.com/docs/editor/debugging), you can run the target with the following configuration: ```json theme={null} { "name": "target-salesforce", "type": "python", "request": "launch", "justMyCode": false, "program": "../target_salesforce/target.py", "cwd": "ABSOLUTE_PATH_TO_TARGET_DIRECTORY/target-salesforce/.secrets", "args": ["<", "data.singer", "--config", "config.json", ">", "state.json"], "console": "integratedTerminal", "python": "ABSOLUTE_PATH_TO_TARGET_DIRECTORY/target-salesforce/.venv/bin/python" } ```