mirror of
https://github.com/discourse/discourse-ai.git
synced 2025-03-01 23:09:19 +00:00
f9d7d7f5f0
* DEV: AI bot migration to the Llm pattern. We added tool and conversation context support to the Llm service in discourse-ai#366, meaning we met all the conditions to migrate this module. This PR migrates to the new pattern, meaning adding a new bot now requires minimal effort as long as the service supports it. On top of this, we introduce the concept of a "Playground" to separate the PM-specific bits from the completion, allowing us to use the bot in other contexts like chat in the future. Commands are called tools, and we simplified all the placeholder logic to perform updates in a single place, making the flow more one-wayish. * Followup fixes based on testing * Cleanup unused inference code * FIX: text-based tools could be in the middle of a sentence * GPT-4-turbo support * Use new LLM API
68 lines
2.1 KiB
Ruby
68 lines
2.1 KiB
Ruby
#frozen_string_literal: true
|
|
|
|
module DiscourseAi
|
|
module AiBot
|
|
module Personas
|
|
class SqlHelper < Persona
|
|
def self.schema
|
|
return @schema if defined?(@schema)
|
|
|
|
tables = Hash.new
|
|
priority_tables = %w[posts topics notifications users user_actions user_emails]
|
|
|
|
DB.query(<<~SQL).each { |row| (tables[row.table_name] ||= []) << row.column_name }
|
|
select table_name, column_name from information_schema.columns
|
|
where table_schema = 'public'
|
|
order by table_name
|
|
SQL
|
|
|
|
schema = +(priority_tables.map { |name| "#{name}(#{tables[name].join(",")})" }.join("\n"))
|
|
|
|
schema << "\nOther tables (schema redacted, available on request): "
|
|
tables.each do |table_name, _|
|
|
next if priority_tables.include?(table_name)
|
|
schema << "#{table_name} "
|
|
end
|
|
|
|
@schema = schema
|
|
end
|
|
|
|
def tools
|
|
[Tools::DbSchema]
|
|
end
|
|
|
|
def system_prompt
|
|
<<~PROMPT
|
|
You are a PostgreSQL expert.
|
|
- You understand and generate Discourse Markdown but specialize in creating queries.
|
|
- You live in a Discourse Forum Message.
|
|
- The schema in your training set MAY be out of date.
|
|
- When generating SQL NEVER end SQL samples with a semicolon (;).
|
|
- When generating SQL always use ```sql markdown code blocks.
|
|
- Always format SQL in a highly readable format.
|
|
|
|
Eg:
|
|
|
|
```sql
|
|
select 1 from table
|
|
```
|
|
|
|
The user_actions tables stores likes (action_type 1).
|
|
the topics table stores private/personal messages it uses archetype private_message for them.
|
|
notification_level can be: {muted: 0, regular: 1, tracking: 2, watching: 3, watching_first_post: 4}.
|
|
bookmarkable_type can be: Post,Topic,ChatMessage and more
|
|
|
|
Current time is: {time}
|
|
|
|
|
|
The current schema for the current DB is:
|
|
{{
|
|
#{self.class.schema}
|
|
}}
|
|
PROMPT
|
|
end
|
|
end
|
|
end
|
|
end
|
|
end
|