This repository contains the FreeScout + ChatGPT Integration Module, which connects FreeScout with the powerful language models of ChatGPT by OpenAI via their API using your own API key. This integration enables the generation of AI-based responses for incoming messages, providing a more efficient and intelligent support system for your FreeScout team.
If this module saves you time, helps your clients, or helps you do better work, I’d appreciate it.
- Generate ChatGPT AI-based responses for each incoming message, then edit/polish as needed
- Utilize the powerful ChatGPT language models to improve support efficiency
- Customizable prompt message to set ChatGPT's role (e.g., support agent, sales manager, etc.), associate it with your brand with website links, and provide additional training information and context as desired
- Edit the prompt per reply using the Edit Prompt button on the Reply page, to add message-specific info to send with the request for a more detailed reply.
To use this module, you will need an API key for ChatGPT, which can be obtained from the OpenAI platform at https://platform.openai.com/account/api-keys.
- Upload the FreeScoutGPT Module ZIP file to your FreeScout Modules folder, rename it to ensure the module's folder is named "FreeScoutGPT".
- Go to each mailbox you want it enabled on, and add your ChatGPT API key to the module's configuration page.
- Set a good prompt to describe it's role, and add some training info, website and KB links, etc for ChatGPT to consider using in it's answers.
- Use the Models drop-down to choose from the various models available to your API Key.
- Be sure to check the Enable switch at the top.
- On a conversation page, click the drop-down in the upper right near the message time and choose "Generate answer (GPT)".
- Then you can copy that answer, click Reply, and paste it into your message, and edit/polish it as desired.
- On the Reply screen, there is an Edit Prompt button, which you can use to rework the prompt and customize it for this reply, and then submit it to generate a reply using the additional information entered. Note that this prompt is not saved, it is unique to this reply.
v1.0.24: No longer unintentionally filters out the new gpt-4.1 models from the Settings model drop-down, which were released April 14, 2025. Cheaper, faster, smarter, I'd recommend using gpt-4.1 over gpt-4o now. ( See: https://openai.com/index/gpt-4-1/ )
v1.0.23: The OpenAI models drop-down on the Settings page is dynamically populated based on the chat completion models available to your API key.
v1.0.22: Change prompt roles based on which model is being used, reasoning models don't support system or admin roles.
v1.0.21: Added auto-update code similar to native modules, and FreeScout was updated to support renaming the folders if need be.
v1.0.20: First release under new forked version, fixed Javascript errors, removed old update code, fixed Reply edit prompt button.
Edit prompt modal window displays current prompt, add message-specific info here and send it to get a generated reply
Based on tiered access frequently changing the models available, the Settings drop-down now dynamically populates the model choices based on your API Key! This gives you access to more models, and keeps the model choices constantly up-to-date without needing to re-code the settings page and release frequent module updates. Note that it caches the models API results for 10 minutes, to avoid hitting the API too frequently. Clearing the FreeScout cache should clear the module's model cache as well, I believe.
OpenAI offers a range of AI language models, each with distinct capabilities and pricing structures. Below is an overview of the models available for chat completions within FreeScoutGPT, with pricing details (which may change and be outdated).
Visit https://platform.openai.com/docs/pricing for the latest pricing.
Be sure to visit https://platform.openai.com/settings/organization/limits to set budget alerts and budget limits, so you don't get any huge surprising invoices/charges.
gpt-4o-mini: Introduced as a cost-efficient alternative, 4o-mini balances performance with affordability. It offers a 128K context window and multimodal capabilities, including text and vision processing. This model replaces gpt-3.5-turbo. Pricing is set at $0.15 per million input tokens and $0.60 per million output tokens. Least expensive and fastest performance with decent quality results.
gpt-4o: GPT-4o (“o” for “omni”) is a versatile, high-intelligence flagship model, ideal for complex tasks requiring high accuracy and advanced capabilities. Less expensive but almost as good in quality and performance to chatgpt-4o-latest, its priced at $2.50 per million input tokens, and $10 per million output tokens.
chatgpt-4o-latest: OpenAI’s advanced chat model, designed to enhance pattern recognition and generate creative insights. It’s particularly adept at tasks requiring high emotional intelligence and creativity, such as writing and brainstorming. However, its advanced capabilities come with a higher cost, priced at $5 per million input tokens, and $15 per million output tokens.
There are new reasoning models as well, including o1, o1-preview, o1-mini, and o3-mini. o1 and o1-preview are priced at $15 per million input tokens, and $60 per million output tokens. o3-mini and o1-mini are both priced at $1.10 per million input tokens, and $4.40 per million output tokens. Note that o1 and o3-mini are being rolled out to Tier 1 API users slowly, and should be used instead of o1-preview and o1-mini if you do have access, better results and same costs. More info on API access to o1 and o3-mini is here: https://help.openai.com/en/articles/10362446-api-access-to-o1-and-o3-mini
A very nice comparison of the models: https://platform.openai.com/docs/models/compare
For more details on the models, visit: https://platform.openai.com/docs/models
Depending on your API tier, you will have access to only certain models. Be aware that currently the free tier only has access to gpt-4o-mini, while most of us will be in Tier 1, and have access to the following chat completion models: gpt-4o, chatgpt-4o-latest, gpt-4o-mini, o1-preview, and o1-mini, and some Tier 1 users now also have access to o1 and o3-mini as they roll them out.
You will gain Tier 2 access once you've spent $50, but it has the same access, with increased request per minute rate limits.
Tier 3 is gained when you spent a total of $100, and it gains access to o1, and o3-mini. Tier 4 and 5 gain increased rate limits to the same models.
To view your API Tier, view the Current Tier section at the bottom of this page: https://platform.openai.com/settings/organization/limits
For more information on the tiers and the rate limits, visit: https://platform.openai.com/docs/guides/rate-limits?tier=tier-one
The default limit of 1024 tokens is about 750 words, and that would include the prompt and generated AI output. If you have more prompt details and training information, and longer messages to send out, then you may want to increase the token limit.
Rough Breakdown: • 1 token ≈ 4 characters (including spaces & punctuation) • 100 tokens ≈ 75 words • 1024 tokens ≈ 750 words
For larger prompts and messages, and larger responses, 2048 or 4096 tokens may work for a better balance of cost and performance. Decrease the token limit for shorter, lower-cost queries to avoid unnecessary expense.
Use OpenAI's Tokenizer to test or preview what your prompt and a typical email represents in input tokens: https://platform.openai.com/tokenizer
This is an updated version of the FreeScout + ChatGPT Integration Module, and we appreciate any feedback, suggestions, or contributions to help improve the module. Please feel free to open issues or submit pull requests on GitHub.
Together, we can make this integration a valuable addition to the FreeScout ecosystem and enhance the capabilities of helpdesk software for the entire community.
If this module saves you time, helps your clients, or helps you do better work, I’d appreciate it.
This module uses the ChatGPT API, developed by OpenAI, and is not officially affiliated with or endorsed by OpenAI. For more details, see OpenAI’s Terms of Service.