gpt integration with slack
- if @ in a channel, reply as a thread
- if send message in DM, reply as a thread
- if chat in a thread, using whole thread as context and reply in the thread
# edit .env
SLACK_TOKEN=<token>
SLACK_REPLY_CHANNEL_TOKEN=<bot_token>
GPT_ORG=<gpt_org_id>
GPT_KEY=<gpt_key>
# run index.py
flask --app index.py run --host 127.0.0.1 --port 8080
# wrap it with nginx to serve https at 443
# fill https://<domain> in slack config and verify the url
bot scope:
app_mentions:read: if you want the bot be @ and it can replychat:write: if you want to enable DM botim:historympim:history
In the code
... messages=[{ ... }, ...]
you need to add first message as a prompt like:
messages = [{ role: 'system', content: prompt } ...]
besides messages, add keys:
messages: [...],
functions: [{ name: 'jsonify', parameters: param }],
function_call: { name: 'jsonify' }
you need to define json format in param like:
type: 'object',
properties: {
company_entities: {
type: 'array',
items: {
type: 'object',
properties: {
name: {
type: 'string',
minLength: 2,
maxLength: 100,
description: 'the name of entity',
},
},
},
description: 'all named entities extracted from the input text',
},
},
and in your prompt you can ask gpt to call jsonify to parse result.
then you can get formated result at for example .choices[0].message.function_call.arguments