Can you share some examples of edge cases?
1 Like
Sure I plan to do another round next week, you can try the ruby script in the repo I shared
@simon I think the real value is understanding that, at least for the foreseeable future, this will never just get something right the first time. But if you know what you want, you can guide it there like a relentless intern that can go and do the grunt work.
So, without having any current knowledge of SQL beyond remembering select from where, and knowing what I wanted my end result to be, I got it to build the query I wanted just by having a conversation with it on the side, and not really taking away from me doing my day job. It really is like having a free personal assistant/intern that I just need to keep guiding in the right direction.
First and foremost, hereās my final query. I wanted a query that would return the top 100 users with the most total likes, and give me their user, username, count of total links, and a link to their one post that had the single most likes. I did not/do not know how to get this at all, and frankly donāt even know where to begin. When I want something like this, I go bother one of my engineers. I was able to not bother them, not slow down from my own work really, and still guide/instruct ChatGPT to get done what I needed done.
Final query:
WITH Most_Liked_Posts AS (
SELECT
p.user_id,
p.topic_id,
p.post_number,
ROW_NUMBER() OVER (PARTITION BY p.user_id ORDER BY likes_count DESC) AS row_number
FROM (
SELECT
p.user_id,
p.topic_id,
p.post_number,
COUNT(l.id) AS likes_count
FROM
posts p
LEFT JOIN post_actions l ON p.id = l.post_id AND l.post_action_type_id = 2
WHERE
p.user_id NOT IN (SELECT id FROM users WHERE username = 'codey')
GROUP BY
p.user_id,
p.topic_id,
p.post_number
) p
),
User_Likes AS (
SELECT
u.id AS user_id,
COUNT(pa.id) AS total_likes
FROM
users u
LEFT JOIN posts p ON u.id = p.user_id
LEFT JOIN post_actions pa ON p.id = pa.post_id AND pa.post_action_type_id = 2
WHERE
u.username != 'codey'
GROUP BY
u.id
)
SELECT
ul.user_id,
u.username AS username,
ul.total_likes,
'<a href="/discuss/t/' || mlp.topic_id || '/' || mlp.post_number || '">' || 'Link to most-liked post' || '</a>' AS html$post
FROM
User_Likes ul
JOIN users u ON ul.user_id = u.id
LEFT JOIN Most_Liked_Posts mlp ON ul.user_id = mlp.user_id AND mlp.row_number = 1
ORDER BY
ul.total_likes DESC
LIMIT 100
Instructions in my ChatGPT account before starting the prompt:
I am only going to ask you quested related to PostgreSQL, and I only want you to respond with relevant SQL queries.
These queries are all related to databases in my community platform, Discourse.
Respond with SQL queries, and explanations of the queries. Do not use any semicolons to end SQL statements as they are not needed.
Follow my entire conversation on ChatGPT building this query here:
https://chat.openai.com/share/6a62ed59-b993-4306-8793-9aaef76f03e9
3 Likes
@sam with a complete schema like the above (again, sorry, I donāt know where to actually get this and/or get it in this format) and using langchain and a vector database to process the documents correctly before sending them into ChatGPT, plus any docs you may have on using the Data Explorerā¦Iām pretty confident this process would be pretty close to magic.
I tried a vector db and feeding it similar examples, it steers it too hard, we are probably going to need 1000 very close examples to be magic
1 Like
Iāll have my engineer try it as well as weāve built a sort of āfactoryā to produce these pretty quickly.
Is there somewhere that I can get a complete and exhaustive document of all db schemas in this format?
# == Schema Information
#
# Table name: application_requests
#
# id :integer not null, primary key
# date :date not null
# req_type :integer not null ("http_total"=>0,"http_2xx"=>1,"http_background"=>2,"http_3xx"=>3,"http_4xx"=>4,"http_5xx"=>5,"page_view_crawler"=>6,"page_view_logged_in"=>7,"page_view_anon"=>8,"page_view_logged_in_mobile"=>9,"page_view_anon_mobile"=>10,"api"=>11,"user_api"=>12)
# count :integer default(0), not null
#
# Table name: users
#
# id :integer not null, primary key
# username :string(60) not null
# created_at :datetime not null
# updated_at :datetime not null
# name :string (the user's real name)
# last_posted_at :datetime
# active :boolean default(FALSE), not null
# username_lower :string(60) not null
# last_seen_at :datetime
# admin :boolean default(FALSE), not null
# trust_level :integer not null
# approved :boolean default(FALSE), not null
# approved_by_id :integer
# approved_at :datetime
# previous_visit_at :datetime
# suspended_at :datetime
# suspended_till :datetime
# date_of_birth :date
# ip_address :inet
# moderator :boolean default(FALSE)
# title :string
# locale :string(10)
# primary_group_id :integer
# registration_ip_address :inet
# staged :boolean default(FALSE), not null
# first_seen_at :datetime
# silenced_till :datetime
1 Like
That format is super wasteful token wise, but you can just query it out of the pg schema tables in data explorer 
2 Likes
Haha okay Iāll have him check it out. Like I said, Iām not the one to take this much further. If I were, Iād probably not be asking these questions!
1 Like
I really want to get something working, but this is a very hard problem
3 Likes
This is just an idea, I have not tried this.
The queries needed IMO break down into uses based on duty
- moderator
- admin
- developer
As such the tables needed can be grouped into ever growing sets with the smallest set being the tables needed by a moderator.
Now, much of the data needed by a moderator will be based on common joins of tables with specific columns being needed, thus a view.
If the common views are used instead of the entire schema, then hopefully many of the prompts would only need to pass along the views and not the entire schema and thus make it much easier for the LLM to generate a possible solution.
HTH
For the much harder queries, odds are if you know enough to need such a query you know enough to build the query.
1 Like
Iāve tried that. The results are all over the place. An interesting experiment is to provide GPT-3.5 with a minimal annotated version of the Discourse database schema just to test its SQL ability. I realize this is inefficient in terms of tokens, but itās readable:
Minimal schema
# == Schema Information
#
# Table name: users
#
# id :integer not null, primary key
# username :string(60) not null
# created_at :datetime not null
#
# Table name: groups
#
# id :integer not null, primary key
# name :string not null
# created_at :datetime not null
#
# Table name: group_users
#
# id :integer not null, primary key
# group_id :integer not null
# user_id :integer not null
#
# Table name: posts
#
# id :integer not null, primary key
# user_id :integer
# topic_id :integer not null
# deleted_at :datetime (The application "soft deletes" posts. When a post is deleted, its `deleted_at` property is set to a :datetime. Unless explicity asked to return deleted posts, make sure the `deleted_at` column is `NOT NULL` when writing queries that are asking for data related to posts.)
#
# Table name: topics
#
# id :integer not null, primary key
# title :string not null
# category_id :integer
# created_at :datetime not null
# user_id :integer (the id of the user who created the topic)
# deleted_at :datetime (The application "soft deletes" topics. When a topic is deleted, its `deleted_at` property is set to a :datetime. Unless explicity asked to return deleted topics, make sure the `deleted_at` column is `NOT NULL` when writing queries that are asking for data related to topics.)
#
# Table name: categories
#
# id :integer not null, primary key
# name :string(50) not null
So far, the results Iām getting are impressive. This gets me thinking that the problem could be solved just by providing an annotated schema in the prompt.
For fun, I asked it what kinds of queries I could expect it to struggle with:
GPT-3.5 limitations
- Hierarchical Queries: Queries involving hierarchical data structures like recursive relationships can be complex. For example, querying for parent-child relationships within a category hierarchy.
- Temporal Queries: Queries involving time-based data, such as calculating historical trends or analyzing data over specific time intervals, might require more intricate SQL constructs.
- Window Functions: While Iāve used window functions in previous queries, more advanced uses of window functions for complex analytical tasks might be challenging.
- Geospatial Queries: Spatial databases and querying based on geographical data, like finding nearby locations or complex spatial relationships, might be complex.
- Complex Aggregations: Advanced aggregations involving multiple levels of grouping and calculations might require more intricate SQL knowledge.
- Optimization Queries: Writing highly optimized queries that involve indexes, query tuning, and performance improvements can be intricate.
- Custom Algorithms: Implementing custom algorithms or logic within a SQL query might be challenging due to the complexity of the logic involved.
- Advanced JOINs: Complex join conditions or combining multiple joins might require careful consideration and possibly more complex SQL constructs.
- Subqueries and Correlated Subqueries: While I can handle subqueries, extremely complex or deeply nested subqueries might be more challenging.
- Pivoting and Unpivoting: Transforming data between wide and tall formats (pivoting and unpivoting) can become complex in certain scenarios.
Where Iām finding problems is with attempting to disambiguate the full database schema. For example, finding a way of annotating the user_actions table. Just providing definitions of its action_type codes isnāt enough. It starts guessing about user_id, target_user_id, and acting_user_id.
The most commonly asked for queries donāt make use of most of the tables and columns in the database. If AI is added to the Data Explorer, it might be worth looking at having ābasicā and āadvancedā modes. Basic mode could supply a prompt that covers most use cases. Advanced mode could allow users to select what information is set in the prompt.
It might be interesting to work backwards from a few requests for queries on meta, to see what would need to be supplied to the prompt in order to have GPT-3.5 successfully create the query.
3 Likes
Potentially a langchain approach where we first ask gpt to determine which tables are relevant followed by a second phase where we generate the sql may help
3 Likes
Our own implementation for our product currently uses langchain. Weāve actually built a bit of a reusable factory, Iām going to have my lead engineer try this out sometime soon.
Like I said though, Iām pretty happy with the results so far. Itās like having an assistant to run errands for meājust needs to make a few trips, but still saves me a ton of time and money as is right now.
2 Likes
FYI
2 Likes
This is super addictive. Based on the āLLMs and SQLā blog post, and a bit of trial and error, I created this prompt that contains a partial description of the Discourse database:
Discourse database prompt
The text between the /* Discourse database documentation start */ and /* Discourse database documentation end */ comments contains details about the Discourse forum application's PostgreSQL database.
All tables and columns are outlined in the `CREATE TABLE` statements. Take note of the sample queries that follow each of the `CREATE TABLE` statements. Some additional important details are contained in
inline (`-- --`) and multi-line (`/* */`) comments. After having sent you this information, I will ask you to write some queries that are to be run by the Discourse Data Explorer plugin. All of the tables and columns
required to write these queries are in the `CREATE TABLE` statements that I have sent you.
/* Discourse database documentation start */
CREATE TABLE users (
id integer NOT NULL, -- the application has the concept of 'real' users. A 'real' user is a user with an id > 0 --
username character varying(60) NOT NULL,
created_at timestamp without time zone NOT NULL,
updated_at timestamp without time zone NOT NULL,
name character varying,
seen_notification_id integer DEFAULT 0 NOT NULL,
last_posted_at timestamp without time zone,
password_hash character varying(64),
salt character varying(32),
active boolean DEFAULT false NOT NULL,
username_lower character varying(60) NOT NULL,
last_seen_at timestamp without time zone,
admin boolean DEFAULT false NOT NULL,
last_emailed_at timestamp without time zone,
trust_level integer NOT NULL,
approved boolean DEFAULT false NOT NULL,
approved_by_id integer,
approved_at timestamp without time zone,
previous_visit_at timestamp without time zone,
suspended_at timestamp without time zone,
suspended_till timestamp without time zone,
date_of_birth date,
views integer DEFAULT 0 NOT NULL,
flag_level integer DEFAULT 0 NOT NULL,
ip_address inet,
moderator boolean DEFAULT false,
title character varying,
uploaded_avatar_id integer,
locale character varying(10),
primary_group_id integer,
registration_ip_address inet,
staged boolean DEFAULT false NOT NULL,
first_seen_at timestamp without time zone,
silenced_till timestamp without time zone,
group_locked_trust_level integer,
manual_locked_trust_level integer,
secure_identifier character varying,
flair_group_id integer,
last_seen_reviewable_id integer
);
SELECT * FROM users WHERE id = 1 OR id = 2 OR id = 121;
id | username | created_at | updated_at | name | seen_notification_id | last_posted_at | password_hash | salt | active | username_lower | last_seen_at | admin | last_emailed_at | trust_level | approved | approved_by_id | approved_at | previous_visit_at | suspended_at | suspended_till | date_of_birth | views | flag_level | ip_address | moderator | title | uploaded_avatar_id | locale | primary_group_id | registration_ip_address | staged | first_seen_at | silenced_till | group_locked_trust_level | manual_locked_trust_level | secure_identifier | flair_group_id | last_seen_reviewable_id | password_algorithm
-----+----------+----------------------------+----------------------------+--------------+----------------------+----------------------------+------------------------------------------------------------------+----------------------------------+--------+----------------+----------------------------+-------+----------------------------+-------------+----------+----------------+----------------------------+----------------------------+----------------------------+-------------------------+---------------+-------+------------+------------+-----------+------------+--------------------+--------+------------------+-------------------------+--------+----------------------------+---------------+--------------------------+---------------------------+------------------------------------------+----------------+-------------------------+------------------------------
1 | scossar | 2019-04-26 22:59:44.685893 | 2023-08-14 04:40:20.823438 | Simon Cossar | 56395 | 2023-08-14 04:08:43.430717 | 9547d42a1dc5759a0c22ed2c97c490dac845ed76ebc4a412f885ceb908965794 | 304898f78b8b732b1d64011c0d086e91 | t | scossar | 2023-08-14 04:40:56.769353 | t | 2023-08-14 04:33:46.44485 | 3 | t | -1 | 2020-09-22 19:54:41.05418 | 2023-08-13 22:35:00.020816 | | | 1904-02-14 | 0 | 0 | ::1 | t | Member | 747 | | | | f | 2019-04-26 23:10:43.250255 | | | 3 | | | 432 | $pbkdf2-sha256$i=64000,l=32$
2 | sally | 2019-04-26 23:15:47.859691 | 2023-08-14 04:40:56.831344 | | 56396 | 2023-08-14 04:11:37.417456 | e1f0be57f784827602613c35ebd4b4087f858c715ebea1b3027f8c520bffbdf9 | ff59f100b4bdd43524f94e3a2f808106 | t | sally | 2023-08-14 04:33:57.054779 | f | 2023-08-14 04:33:06.727322 | 2 | t | -1 | 2020-05-19 19:35:15.79381 | 2023-08-13 22:08:05.099486 | | | | 0 | 0 | 127.0.0.1 | t | Regular | 22 | en | 49 | 127.0.0.1 | f | 2019-04-26 23:16:58.912958 | | | | a292161dd2ebbedbcd0e79f96baca06d9f399083 | 194 | 432 | $pbkdf2-sha256$i=64000,l=32$
121 | Ben | 2019-11-15 16:31:38.216013 | 2023-08-14 04:40:41.907605 | | 56314 | 2023-07-07 20:48:33.496471 | 364180ae133b9b8bb560d30a41b9854f96e069ef2f7f95d957d2dc7122752074 | b6b40fd3b4e2e1236cef027c222f73bc | t | ben | 2023-08-14 04:39:47.42192 | f | 2023-08-14 04:30:26.764459 | 2 | t | 1 | 2019-11-15 16:31:38.089553 | 2023-07-22 02:14:01.540478 | 2022-05-05 17:39:02.632952 | 2022-05-06 17:38:55.054 | | 0 | 0 | 127.0.0.1 | f | Prime Four | | en | 196 | 127.0.0.1 | f | 2019-11-15 16:31:38.714185 | | | | | 196 | | $pbkdf2-sha256$i=64000,l=32$
CREATE TABLE groups (
id integer NOT NULL,
name character varying NOT NULL,
created_at timestamp without time zone NOT NULL,
updated_at timestamp without time zone NOT NULL,
automatic boolean DEFAULT false NOT NULL,
user_count integer DEFAULT 0 NOT NULL,
automatic_membership_email_domains text,
primary_group boolean DEFAULT false NOT NULL,
title character varying,
grant_trust_level integer,
incoming_email character varying,
has_messages boolean DEFAULT false NOT NULL,
flair_url character varying,
flair_bg_color character varying,
flair_color character varying,
bio_raw text,
bio_cooked text,
allow_membership_requests boolean DEFAULT false NOT NULL,
full_name character varying,
default_notification_level integer DEFAULT 3 NOT NULL,
visibility_level integer DEFAULT 0 NOT NULL,
public_exit boolean DEFAULT false NOT NULL,
public_admission boolean DEFAULT false NOT NULL,
membership_request_template text,
messageable_level integer DEFAULT 0,
mentionable_level integer DEFAULT 0,
members_visibility_level integer DEFAULT 0 NOT NULL,
publish_read_state boolean DEFAULT false NOT NULL,
flair_icon character varying,
flair_upload_id integer,
smtp_server character varying,
smtp_port integer,
smtp_ssl boolean,
imap_server character varying,
imap_port integer,
imap_ssl boolean,
imap_mailbox_name character varying DEFAULT ''::character varying NOT NULL,
imap_uid_validity integer DEFAULT 0 NOT NULL,
imap_last_uid integer DEFAULT 0 NOT NULL,
email_username character varying,
email_password character varying,
imap_last_error text,
imap_old_emails integer,
imap_new_emails integer,
allow_unknown_sender_topic_replies boolean DEFAULT false NOT NULL,
smtp_enabled boolean DEFAULT false,
smtp_updated_at timestamp without time zone,
smtp_updated_by_id integer,
imap_enabled boolean DEFAULT false,
imap_updated_at timestamp without time zone,
imap_updated_by_id integer,
assignable_level integer DEFAULT 0 NOT NULL,
email_from_alias character varying
); -- users who have either 'admin' or 'moderator' status are added to the automatic "staff" group --
SELECT * FROM groups WHERE id = 1 OR id = 11 OR id = 49;
id | name | created_at | updated_at | automatic | user_count | automatic_membership_email_domains | primary_group | title | grant_trust_level | incoming_email | has_messages | flair_bg_color | flair_color | bio_raw | bio_cooked | allow_membership_requests | full_name | default_notification_level | visibility_level | public_exit | public_admission | membership_request_template | messageable_level | mentionable_level | members_visibility_level | publish_read_state | flair_icon | flair_upload_id | smtp_server | smtp_port | smtp_ssl | imap_server | imap_port | imap_ssl | imap_mailbox_name | imap_uid_validity | imap_last_uid | email_username | email_password | imap_last_error | imap_old_emails | imap_new_emails | allow_unknown_sender_topic_replies | smtp_enabled | smtp_updated_at | smtp_updated_by_id | imap_enabled | imap_updated_at | imap_updated_by_id | assignable_level | email_from_alias
----+---------------+----------------------------+----------------------------+-----------+------------+------------------------------------+---------------+------------+-------------------+----------------+--------------+----------------+-------------+--------------------+---------------------------+---------------------------+---------------------------+----------------------------+------------------+-------------+------------------+-----------------------------+-------------------+-------------------+--------------------------+--------------------+------------+-----------------+-------------+-----------+----------+-------------+-----------+----------+-------------------+-------------------+---------------+----------------+----------------+-----------------+-----------------+-----------------+------------------------------------+--------------+---------------------------+--------------------+--------------+-----------------+--------------------+------------------+------------------
1 | admins | 2019-04-26 22:58:35.997964 | 2021-08-05 19:11:22.699825 | t | 1 | | f | | | | t | | | | | f | | 3 | 1 | f | f | | 99 | 0 | 0 | f | | | | | | | | | | 0 | 0 | | | | | | f | f | | | f | | | 0 |
11 | trust_level_1 | 2019-04-26 22:58:36.033238 | 2021-10-05 19:54:51.043121 | t | 116 | | f | | | | t | | | | | f | | 3 | 1 | f | f | | 0 | 0 | 0 | f | | | | | | | | | | 0 | 0 | | | | | | f | f | | | f | | | 0 |
49 | eurorack | 2019-10-03 17:28:42.323203 | 2022-08-16 19:54:09.223307 | f | 84 | example.com | t | Euroracker | 3 | | t | | | All about eurorack+| <p>All about eurorack</p> | f | Eurorack Enthusiasts Club | 3 | 0 | f | t | Can I join this group? | 99 | 99 | 0 | t | | | | | | | | | | 0 | 0 | | | | | | f | f | 2022-02-11 23:24:28.76631 | 1 | f | | | 0 |
/* group_users joins the groups and users tables */
CREATE TABLE group_users (
id integer NOT NULL,
group_id integer NOT NULL,
user_id integer NOT NULL,
created_at timestamp without time zone NOT NULL,
updated_at timestamp without time zone NOT NULL,
owner boolean DEFAULT false NOT NULL,
notification_level integer DEFAULT 2 NOT NULL,
first_unread_pm_at timestamp without time zone DEFAULT CURRENT_TIMESTAMP NOT NULL
);
SELECT * FROM group_users WHERE id = 13 OR id = 8219 OR id = 9137;
id | group_id | user_id | created_at | updated_at | owner | notification_level | first_unread_pm_at
------+----------+---------+----------------------------+----------------------------+-------+--------------------+----------------------------
13 | 3 | 1 | 2019-04-26 22:59:47.828533 | 2019-04-26 22:59:47.828533 | f | 2 | 2023-08-14 01:14:54.229593
8219 | 13 | 121 | 2022-04-21 08:15:47.946036 | 2022-04-21 08:15:47.946036 | f | 2 | 2023-07-05 06:49:04.48265
9137 | 49 | 2 | 2022-09-08 17:34:39.290504 | 2022-09-08 17:34:39.290504 | t | 3 | 2020-11-21 02:40:15.868728
CREATE TABLE posts (
id integer NOT NULL,
user_id integer,
topic_id integer NOT NULL,
post_number integer NOT NULL,
raw text NOT NULL,
cooked text NOT NULL,
created_at timestamp without time zone NOT NULL,
updated_at timestamp without time zone NOT NULL,
reply_to_post_number integer,
reply_count integer DEFAULT 0 NOT NULL,
quote_count integer DEFAULT 0 NOT NULL,
deleted_at timestamp without time zone, -- the application only "soft deletes" posts and topics. Unless explicitly asked to return details about deleted posts or topics, always check that `deleted_at IS NULL` when writing queries related to posts or topics. --
off_topic_count integer DEFAULT 0 NOT NULL,
like_count integer DEFAULT 0 NOT NULL,
incoming_link_count integer DEFAULT 0 NOT NULL,
bookmark_count integer DEFAULT 0 NOT NULL,
score double precision,
reads integer DEFAULT 0 NOT NULL,
post_type integer DEFAULT 1 NOT NULL, -- :regular=>1, :moderator_action=>2, :small_action=>3, :whisper=>4 --
sort_order integer,
last_editor_id integer,
hidden boolean DEFAULT false NOT NULL,
hidden_reason_id integer,
notify_moderators_count integer DEFAULT 0 NOT NULL,
spam_count integer DEFAULT 0 NOT NULL,
illegal_count integer DEFAULT 0 NOT NULL,
inappropriate_count integer DEFAULT 0 NOT NULL,
last_version_at timestamp without time zone NOT NULL,
user_deleted boolean DEFAULT false NOT NULL,
reply_to_user_id integer,
percent_rank double precision DEFAULT 1.0,
notify_user_count integer DEFAULT 0 NOT NULL,
like_score integer DEFAULT 0 NOT NULL,
deleted_by_id integer,
edit_reason character varying,
word_count integer,
version integer DEFAULT 1 NOT NULL,
cook_method integer DEFAULT 1 NOT NULL,
wiki boolean DEFAULT false NOT NULL,
baked_at timestamp without time zone,
baked_version integer,
hidden_at timestamp without time zone,
self_edits integer DEFAULT 0 NOT NULL,
reply_quoted boolean DEFAULT false NOT NULL,
via_email boolean DEFAULT false NOT NULL,
raw_email text,
public_version integer DEFAULT 1 NOT NULL,
action_code character varying,
locked_by_id integer,
image_upload_id bigint
);
SELECT * FROM posts WHERE id = 11094 OR id = 11095 OR id = 11096;
id | user_id | topic_id | post_number | raw | cooked | created_at | updated_at | reply_to_post_number | reply_count | quote_count | deleted_at | off_topic_count | like_count | incoming_link_count | bookmark_count | score | reads | post_type | sort_order | last_editor_id | hidden | hidden_reason_id | notify_moderators_count | spam_count | illegal_count | inappropriate_count | last_version_at | user_deleted | reply_to_user_id | percent_rank | notify_user_count | like_score | deleted_by_id | edit_reason | word_count | version | cook_method | wiki | baked_at | baked_version | hidden_at | self_edits | reply_quoted | via_email | raw_email | public_version | action_code | locked_by_id | image_upload_id | outbound_message_id
-------+---------+----------+-------------+----------------------------------------------+-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+----------------------------+----------------------------+----------------------+-------------+-------------+------------+-----------------+------------+---------------------+----------------+-------+-------+-----------+------------+----------------+--------+------------------+-------------------------+------------+---------------+---------------------+----------------------------+--------------+------------------+-------------------+-------------------+------------+---------------+-------------+------------+---------+-------------+------+----------------------------+---------------+-----------+------------+--------------+-----------+-----------+----------------+-------------+--------------+-----------------+--------------------------------
11094 | 1 | 1852 | 7 | This is an example of a Discourse post. | <p>This is an example of a Discourse post.</p> | 2023-08-14 04:08:43.430717 | 2023-08-14 04:08:43.430717 | | 0 | 0 | | 0 | 0 | 0 | 0 | 0.2 | 1 | 1 | 7 | 1 | f | | 0 | 0 | 0 | 0 | 2023-08-14 04:08:43.441371 | f | | 0.166666666666667 | 0 | 0 | | | 8 | 1 | 1 | f | 2023-08-14 04:08:43.430685 | 2 | | 0 | f | f | | 1 | | | |
11095 | 2 | 10863 | 11 | This is another example of a Discourse post. | <p>This is another example of a Discourse post.</p> | 2023-08-14 04:11:37.417456 | 2023-08-14 04:11:37.417456 | 5 | 0 | 0 | | 0 | 0 | 0 | 0 | 0.4 | 2 | 1 | 11 | 2 | f | | 0 | 0 | 0 | 0 | 2023-08-14 04:11:37.430459 | f | 121 | 0.8 | 0 | 0 | | | 8 | 1 | 1 | f | 2023-08-14 04:11:37.417417 | 2 | | 0 | f | f | | 1 | | | | discourse/post/11095@127.0.0.1
11096 | 121 | 11047 | 2 | Thanks! I hadn't seen that :slight_smile: | <p>Thanks! I hadnāt seen that <img src="//127.0.0.1:4200/images/emoji/twitter/slight_smile.png?v=12" title=":slight_smile:" class="emoji" alt=":slight_smile:" loading="lazy" width="20" height="20"></p> | 2023-08-14 04:13:09.346146 | 2023-08-14 05:17:10.659726 | | 0 | 0 | | 0 | 0 | 0 | 0 | 0.4 | 2 | 1 | 2 | 1 | f | | 0 | 0 | 0 | 0 | 2023-08-14 05:17:10.589855 | f | | 0 | 0 | 0 | | | 7 | 2 | 1 | f | 2023-08-14 05:17:10.659607 | 2 | | 0 | f | f | | 2 | | | | discourse/post/11096@127.0.0.1
CREATE TABLE topics (
id integer NOT NULL,
title character varying NOT NULL,
last_posted_at timestamp without time zone,
created_at timestamp without time zone NOT NULL,
updated_at timestamp without time zone NOT NULL,
views integer DEFAULT 0 NOT NULL,
posts_count integer DEFAULT 0 NOT NULL,
user_id integer, -- user id the id of the user who created the topic --
last_post_user_id integer NOT NULL,
reply_count integer DEFAULT 0 NOT NULL,
featured_user1_id integer,
featured_user2_id integer,
featured_user3_id integer,
deleted_at timestamp without time zone, -- the application only "soft deletes" posts and topics. Unless explicitly asked to return details about deleted posts or topics, always check that `deleted_at IS NULL` when writing queries related to posts or topics. --
highest_post_number integer DEFAULT 0 NOT NULL,
like_count integer DEFAULT 0 NOT NULL,
incoming_link_count integer DEFAULT 0 NOT NULL,
category_id integer,
visible boolean DEFAULT true NOT NULL,
moderator_posts_count integer DEFAULT 0 NOT NULL,
closed boolean DEFAULT false NOT NULL,
archived boolean DEFAULT false NOT NULL,
bumped_at timestamp without time zone NOT NULL,
has_summary boolean DEFAULT false NOT NULL,
archetype character varying DEFAULT 'regular'::character varying NOT NULL, -- archetype can be set to either 'regular' or 'private_message' --
featured_user4_id integer,
notify_moderators_count integer DEFAULT 0 NOT NULL,
spam_count integer DEFAULT 0 NOT NULL,
pinned_at timestamp without time zone,
score double precision,
percent_rank double precision DEFAULT 1.0 NOT NULL,
subtype character varying,
slug character varying,
deleted_by_id integer,
participant_count integer DEFAULT 1,
word_count integer,
excerpt character varying,
pinned_globally boolean DEFAULT false NOT NULL,
pinned_until timestamp without time zone,
fancy_title character varying,
highest_staff_post_number integer DEFAULT 0 NOT NULL,
featured_link character varying,
reviewable_score double precision DEFAULT 0.0 NOT NULL,
image_upload_id bigint,
slow_mode_seconds integer DEFAULT 0 NOT NULL,
bannered_until timestamp without time zone,
external_id character varying,
CONSTRAINT has_category_id CHECK (((category_id IS NOT NULL) OR ((archetype)::text <> 'regular'::text))),
CONSTRAINT pm_has_no_category CHECK (((category_id IS NULL) OR ((archetype)::text <> 'private_message'::text)))
);
SELECT * FROM topics WHERE id = 1852 OR id = 10863 OR id = 11047;
id | title | last_posted_at | created_at | updated_at | views | posts_count | user_id | last_post_user_id | reply_count | featured_user1_id | featured_user2_id | featured_user3_id | deleted_at | highest_post_number | like_count | incoming_link_count | category_id | visible | moderator_posts_count | closed | archived | bumped_at | has_summary | archetype | featured_user4_id | notify_moderators_count | spam_count | pinned_at | score | percent_rank | subtype | slug | deleted_by_id | participant_count | word_count | excerpt | pinned_globally | pinned_until | fancy_title | highest_staff_post_number | featured_link | reviewable_score | image_upload_id | slow_mode_seconds | bannered_until | external_id
-------+-------------------------------------+----------------------------+----------------------------+----------------------------+-------+-------------+---------+-------------------+-------------+-------------------+-------------------+-------------------+------------+---------------------+------------+---------------------+-------------+---------+-----------------------+--------+----------+----------------------------+-------------+-----------------+-------------------+-------------------------+------------+-----------+-------------------+--------------+--------------+-------------------------------------+---------------+-------------------+------------+-----------------------------------------------------------------------------------+-----------------+--------------+-------------------------------------+---------------------------+---------------+------------------+-----------------+-------------------+----------------+-------------
1852 | Ask Me Anything | 2023-08-14 04:08:43.430717 | 2020-05-04 22:11:50.813596 | 2023-08-14 05:25:00.249902 | 2 | 3 | 1 | 1 | 0 | | | | | 7 | 0 | 2 | 3 | t | 4 | f | f | 2023-08-14 04:08:43.430717 | f | regular | | 0 | 0 | | 0.914285714285714 | 1 | | ask-me-anything | | 1 | 157 | The AMA with @sally will be held on May 8th. Start sending in your questions now. | f | | Ask Me Anything | 7 | | 0 | | 0 | |
10863 | Post the last picture on your phone | 2023-08-14 04:11:37.417456 | 2022-01-25 21:42:52.514124 | 2023-08-14 05:22:30.373683 | 18 | 5 | 2 | 2 | 3 | 121 | 1 | | | 11 | 4 | 7 | 14 | t | 0 | f | f | 2023-08-14 04:11:37.417456 | f | regular | | 0 | 0 | | 6.54545454545455 | 1 | | post-the-last-picture-on-your-phone | | 3 | 150 | This is a test. This should go to the review queue | f | | Post the last picture on your phone | 11 | | 70.75 | 445 | 0 | |
11047 | Have you seen this post? | 2023-08-14 04:13:09.346146 | 2022-05-12 22:26:39.280698 | 2023-08-14 05:16:41.106722 | 3 | 2 | 1 | 121 | 0 | | | | | 2 | 0 | 0 | | t | 0 | f | f | 2023-08-14 05:17:10.695009 | f | private_message | | 0 | 0 | | 0.4 | 1 | user_to_user | have-you-seen-this-post | | 2 | 11 | Here is another oneā¦ | f | | Have you seen this post? | 2 | | 0 | | 0 | |
CREATE TABLE categories (
id integer NOT NULL,
name character varying(50) NOT NULL,
color character varying(6) DEFAULT '0088CC'::character varying NOT NULL,
topic_id integer,
topic_count integer DEFAULT 0 NOT NULL,
created_at timestamp without time zone NOT NULL,
updated_at timestamp without time zone NOT NULL,
user_id integer NOT NULL,
topics_year integer DEFAULT 0,
topics_month integer DEFAULT 0,
topics_week integer DEFAULT 0,
slug character varying NOT NULL,
description text,
text_color character varying(6) DEFAULT 'FFFFFF'::character varying NOT NULL,
read_restricted boolean DEFAULT false NOT NULL,
auto_close_hours double precision,
post_count integer DEFAULT 0 NOT NULL,
latest_post_id integer,
latest_topic_id integer,
"position" integer,
parent_category_id integer,
posts_year integer DEFAULT 0,
posts_month integer DEFAULT 0,
posts_week integer DEFAULT 0,
email_in character varying,
email_in_allow_strangers boolean DEFAULT false,
topics_day integer DEFAULT 0,
posts_day integer DEFAULT 0,
allow_badges boolean DEFAULT true NOT NULL,
name_lower character varying(50) NOT NULL,
auto_close_based_on_last_post boolean DEFAULT false,
topic_template text,
contains_messages boolean,
sort_order character varying,
sort_ascending boolean,
uploaded_logo_id integer,
uploaded_background_id integer,
topic_featured_link_allowed boolean DEFAULT true,
all_topics_wiki boolean DEFAULT false NOT NULL,
show_subcategory_list boolean DEFAULT false,
num_featured_topics integer DEFAULT 3,
default_view character varying(50),
subcategory_list_style character varying(50) DEFAULT 'rows_with_featured_topics'::character varying,
default_top_period character varying(20) DEFAULT 'all'::character varying,
mailinglist_mirror boolean DEFAULT false NOT NULL,
minimum_required_tags integer DEFAULT 0 NOT NULL,
navigate_to_first_post_after_read boolean DEFAULT false NOT NULL,
search_priority integer DEFAULT 0,
allow_global_tags boolean DEFAULT false NOT NULL,
reviewable_by_group_id integer,
read_only_banner character varying,
default_list_filter character varying(20) DEFAULT 'all'::character varying,
allow_unlimited_owner_edits_on_first_post boolean DEFAULT false NOT NULL,
default_slow_mode_seconds integer
);
SELECT * FROM categories WHERE id = 3 OR id = 16 OR id = 42;
id | name | color | topic_id | topic_count | created_at | updated_at | user_id | topics_year | topics_month | topics_week | slug | description | text_color | read_restricted | auto_close_hours | post_count | latest_post_id | latest_topic_id | position | parent_category_id | posts_year | posts_month | posts_week | email_in | email_in_allow_strangers | topics_day | posts_day | allow_badges | name_lower | auto_close_based_on_last_post | topic_template | contains_messages | sort_order | sort_ascending | uploaded_logo_id | uploaded_background_id | topic_featured_link_allowed | all_topics_wiki | show_subcategory_list | num_featured_topics | default_view | subcategory_list_style | default_top_period | mailinglist_mirror | minimum_required_tags | navigate_to_first_post_after_read | search_priority | allow_global_tags | reviewable_by_group_id | read_only_banner | default_list_filter | allow_unlimited_owner_edits_on_first_post | default_slow_mode_seconds | uploaded_logo_dark_id
----+------------------+--------+----------+-------------+----------------------------+----------------------------+---------+-------------+--------------+-------------+------------------+-------------------------------------------------------------------------------------------+------------+-----------------+------------------+------------+----------------+-----------------+----------+--------------------+------------+-------------+------------+----------+--------------------------+------------+-----------+--------------+------------------+-------------------------------+----------------+-------------------+------------+----------------+------------------+------------------------+-----------------------------+-----------------+-----------------------+---------------------+--------------+---------------------------+--------------------+--------------------+-----------------------+-----------------------------------+-----------------+-------------------+------------------------+------------------+---------------------+-------------------------------------------+---------------------------+-----------------------
3 | Staff | E45735 | 2 | 16 | 2019-04-26 22:58:38.813759 | 2023-02-16 08:35:51.424024 | -1 | 0 | 0 | 0 | staff | Private category for staff discussions. Topics are only visible to admins and moderators. | FFFFFF | t | | 23 | 11094 | 11250 | 41 | | 1 | 0 | 0 | | f | 0 | 0 | t | staff | f | | | | | | | t | f | f | 3 | | rows_with_featured_topics | all | f | 0 | f | 0 | f | | | all | f | |
16 | customer support | F7941D | 277 | 13 | 2019-07-17 20:18:13.584715 | 2023-08-11 20:50:35.548498 | 1 | 1 | 0 | 0 | customer-support | This description will appear on the categories page. | FFFFFF | f | 720 | 68 | 11087 | 11236 | 0 | | 10 | 7 | 0 | | f | 0 | 0 | f | customer support | f | | | | | 689 | | t | f | f | 4 | | rows_with_featured_topics | all | f | 0 | f | 0 | f | | | all | f | |
42 | eurorack | 0088CC | 584 | 30 | 2019-10-03 17:29:33.782372 | 2023-08-13 21:35:09.514998 | 1 | 1 | 0 | 0 | eurorack | All about Eurorack synths. | FFFFFF | f | | 142 | 11091 | 11183 | 1 | | 15 | 3 | 3 | | f | 0 | 3 | t | eurorack | f | | | posts | | | | t | f | f | 4 | latest | rows_with_featured_topics | all | f | 2 | t | 0 | f | | | all | f | |
CREATE TABLE tags (
id integer NOT NULL,
name character varying NOT NULL,
topic_count integer DEFAULT 0 NOT NULL,
created_at timestamp without time zone NOT NULL,
updated_at timestamp without time zone NOT NULL,
pm_topic_count integer DEFAULT 0 NOT NULL,
target_tag_id integer,
description character varying
);
SELECT * FROM tags WHERE id = 183 OR id = 184 OR id = 185;
id | name | created_at | updated_at | pm_topic_count | target_tag_id | description | public_topic_count | staff_topic_count
-----+-------------+----------------------------+----------------------------+----------------+---------------+-------------+--------------------+-------------------
183 | photos | 2023-08-14 05:22:30.326838 | 2023-08-14 05:22:30.326838 | 0 | | | 1 | 1
184 | meetup | 2023-08-14 05:25:00.227547 | 2023-08-14 05:25:00.227547 | 0 | | | 0 | 1
185 | icebreakers | 2023-08-14 06:18:39.214459 | 2023-08-14 06:18:39.214459 | 0 | | | 1 | 1
CREATE TABLE topic_tags (
id integer NOT NULL,
topic_id integer NOT NULL,
tag_id integer NOT NULL,
created_at timestamp without time zone NOT NULL,
updated_at timestamp without time zone NOT NULL
);
SELECT * FROM topic_tags WHERE id = 1005 OR id = 1006 OR id = 1007;
id | topic_id | tag_id | created_at | updated_at
------+----------+--------+----------------------------+----------------------------
1005 | 10863 | 183 | 2023-08-14 05:22:30.360668 | 2023-08-14 05:22:30.360668
1006 | 1852 | 184 | 2023-08-14 05:25:00.237298 | 2023-08-14 05:25:00.237298
1007 | 10863 | 185 | 2023-08-14 06:18:39.240753 | 2023-08-14 06:18:39.240753
CREATE TABLE user_actions (
id integer NOT NULL,
action_type integer NOT NULL, -- :like=>1 (when a user likes a post), :was_liked=>2 (when a user's post is liked), :new_topic=>4, :reply=>5, :response=>6, :mention=>7, :quote=>9, :edit=>11, :new_private_message=>12, :got_private_message=>13, :solved=>15, :assigned=>16 --
user_id integer NOT NULL,
target_topic_id integer,
target_post_id integer,
target_user_id integer,
acting_user_id integer,
created_at timestamp without time zone NOT NULL,
updated_at timestamp without time zone NOT NULL
);
SELECT * FROM user_actions WHERE id = 19928 OR id = 19929 OR id = 19931;
/*
In the first example below, the user with the id 1 has liked (action_type: 1) the post with the id 11100. Because the user with the id 1 took the action, 1 is set as the value of the acting_user_id column. Because the entry is recording that this user has performed a "like" action, their user id (1) is also set in the entry's user_id column.
Here's an example query that would return the number of times the user with id: 1 has liked other user's posts: `SELECT COUNT(user_id) AS number_of_likes_given FROM user_actions WHERE action_type = 1 AND user_id = 1;`
In the second example below, the user with the id 121 has had their post (id 11100) liked by another user (action_type: 2). The entry's user_id is set to 121 because that's the id of the user who has had their post liked. The entry's acting_user_id column is set to 1 because that is the user who liked the post.
Here's an example query that returns the number of times the user with id 121 has had their posts liked by other users: `SELECT COUNT(user_id) AS number_of_likes_received FROM user_actions WHERE action_type = 2 AND user_id = 121;`
Here's an example query that returns the number of times the user with id 121 has had their posts liked by the user with id: 1: `SELECT COUNT(user_id) AS number_of_likes_received_from_user_1 FROM user_actions WHERE action_type = 2 AND user_id = 121 AND acting_user_id = 1;`
In the third example, the user with the id 121 has been mentioned (action_type: 7) in the post (id 11101) by the user with the id 2. The entry's user_id is set to 121 because that is the id of the user who was mentioned. The entry's acting_user_id is set to 2 because that is the id of the user who created the mention.
Here's an example that returns the number of times the user with id 121 has been mentioned by any users: `SELECT COUNT(user_id) AS number_of_mentions_received FROM user_actions WHERE action_type = 7 AND user_id = 121;`
Here's an example query that returns the number of times the user with id 121 has been mentioned by the user with id 2: `SELECT COUNT(user_id) AS number_of_mentions_received_from_user_2 FROM user_actions WHERE action_type = 7 AND user_id = 121 AND acting_user_id = 2;`
Here's an example query that returns the number of times the user with id 121 has been mentioned in the topic with the id 7282: `SELECT COUNT(user_id) AS number_of_mentions_received_in_topic FROM user_actions WHERE action_type = 7 AND user_id = 121 AND target_topic_id = 7282;`
*/
id | action_type | user_id | target_topic_id | target_post_id | target_user_id | acting_user_id | created_at | updated_at
-------+-------------+---------+-----------------+----------------+----------------+----------------+----------------------------+----------------------------
19928 | 1 | 1 | 7282 | 11100 | | 1 | 2023-08-14 07:47:49.520171 | 2023-08-14 07:47:49.651056
19929 | 2 | 121 | 7282 | 11100 | | 1 | 2023-08-14 07:47:49.520171 | 2023-08-14 07:47:49.671757
19931 | 7 | 121 | 7282 | 11101 | | 2 | 2023-08-14 08:00:05.877537 | 2023-08-14 08:00:05.877537
CREATE TABLE polls (
id bigint NOT NULL,
post_id bigint,
name character varying DEFAULT 'poll'::character varying NOT NULL,
close_at timestamp without time zone,
type integer DEFAULT 0 NOT NULL,
status integer DEFAULT 0 NOT NULL,
results integer DEFAULT 0 NOT NULL,
visibility integer DEFAULT 0 NOT NULL,
min integer,
max integer,
step integer,
anonymous_voters integer,
created_at timestamp without time zone NOT NULL,
updated_at timestamp without time zone NOT NULL,
chart_type integer DEFAULT 0 NOT NULL, -- {"bar"=>0, "pie"=>1} --
groups character varying,
title character varying
);
SELECT * FROM polls WHERE id = 71 OR id = 72 OR id = 74;
id | post_id | name | close_at | type | status | results | visibility | min | max | step | anonymous_voters | created_at | updated_at | chart_type | groups | title
----+---------+------+---------------------+------+--------+---------+------------+-----+-----+------+------------------+----------------------------+----------------------------+------------+---------------+-------------------------------------------
71 | 11097 | poll | 2023-08-20 07:00:00 | 0 | 0 | 0 | 1 | | | | | 2023-08-14 06:38:10.96388 | 2023-08-14 06:38:10.96388 | 0 | trust_level_2 | Who took the best picture?
72 | 11098 | poll | 2023-09-10 07:00:00 | 0 | 0 | 0 | 1 | | | | | 2023-08-14 06:40:36.925762 | 2023-08-14 06:40:36.925762 | 1 | staff | Who should we invite to our next AMA?
74 | 11100 | poll | 2023-08-27 07:00:00 | 1 | 0 | 0 | 0 | 1 | 2 | | | 2023-08-14 06:48:27.498764 | 2023-08-14 06:48:27.498764 | 0 | eurorack | What are your favourite Eurorack modules?
CREATE TABLE poll_options (
id bigint NOT NULL,
poll_id bigint,
digest character varying NOT NULL,
html text NOT NULL,
anonymous_votes integer,
created_at timestamp without time zone NOT NULL,
updated_at timestamp without time zone NOT NULL
);
SELECT * FROM poll_options WHERE id = 271 OR id = 274 OR id = 280;
id | poll_id | digest | html | anonymous_votes | created_at | updated_at
-----+---------+----------------------------------+----------------+-----------------+----------------------------+----------------------------
271 | 71 | 5c6c2c880d86e5a0d9f0924a7ffd9629 | Sally | | 2023-08-14 06:38:10.975263 | 2023-08-14 06:38:10.975263
274 | 72 | 2654933188fb6bd444b6df4cc39fd908 | Dangerous Dave | | 2023-08-14 06:40:36.930354 | 2023-08-14 06:40:36.930354
280 | 74 | f2e5462d97476629719e607dc80a9619 | Maths | | 2023-08-14 06:48:27.502673 | 2023-08-14 06:48:27.502673
CREATE TABLE poll_votes (
poll_id bigint,
poll_option_id bigint,
user_id bigint,
created_at timestamp without time zone NOT NULL,
updated_at timestamp without time zone NOT NULL
);
SELECT * FROM poll_votes WHERE poll_id = 71;
poll_id | poll_option_id | user_id | created_at | updated_at
---------+----------------+---------+----------------------------+----------------------------
71 | 270 | 2 | 2023-08-14 06:59:00.382873 | 2023-08-14 06:59:00.382873
71 | 271 | 121 | 2023-08-14 07:02:57.364892 | 2023-08-14 07:02:57.364892
71 | 271 | 1 | 2023-08-14 07:03:36.464304 | 2023-08-14 07:03:36.464304
/* Discourse database documentation end */
It covers the users, groups, group_users, posts, topics, categories, tags, topic_tags, user_actions, polls, poll_options, and poll_votes tables. Itās currently at 1150 tokens, so it should be able to double its size without causing issues. Note that if you copy it into a ChatGPT chat input, youāll need to paste it into two separate inputs - the character limit on a chat input is less than the token limit for a chat session.
I added fairly detailed comments above the example user_actions queries. With those examples, ChatGPT is doing a good job in answering questions about likes given, likes received, etc. Previously it had been struggling with that. I suspect there are a few tables that would need a similar approach.
After having sent the documentation, the following prompts are helpful:
Example prompts
When I ask you start a query with a āquery period CTEā, I want the query to start with exactly this SQL (including the comment):
--[params]
-- string :query_interval = 1 week
-- date :start_date
-- date :end_date
WITH query_periods AS (
SELECT
generate_series(:start_date, :end_date, :query_interval::interval)::date AS period_start,
(generate_series(:start_date, :end_date, :query_interval::interval)::date + :query_interval::interval - INTERVAL '1 DAY')::date AS period_end
)
The Data Explorer plugin allows paramters to be added to its queries. Parameters that are used in queries need to appear in a comment at the top of the query in this form:
--[params]
--param_type :param_name
The available param types are:
int,bigint,boolean,string,date,time,datetime,double,user_id,post_id,topic_id,category_id,group_id,badge_id,int_list,string_list,user_list
Here is an example:
--[params]
-- string :action_type
An optional default value can be supplied for a parameter. For example:
--[params]
--string :action_type = like
I created the āquery_period CTEā prompt because if I didnāt specifically tell ChatGPT how to create the query_period CTE, it was coming up will all sorts of solutions, some better than others. With the prompt, it adds the exact code I give it, then builds queries on top of it without any problems. Interestingly, when I tried to add details about how to create a āquery period CTEā to the initial documentation that I sent ChatGPT, it would just ignore the instructions. For some reason sending it as a separate prompt makes a difference.
Prompts describing how Discourse āsoft deletesā topics and posts are also helpful. After letting ChatGPT know about the need to check that deleted_at IS NOT NULL for queries related to topics and posts, it consistently applies that to all queries.
Telling ChatGPT to leave off the semicolon at the end of the queries is kind of hopeless. It remembers for a couple of queries, then goes back to adding the semicolon. That seems like a minor detail.
My initial hope of having a perfect query returned from a natural language question was a bit ambitious. It makes mistakes, and so do I. At least in the short term, I think the best way to integrate ChatGPT with the Data Explorer plugin would be to initiate a PM with ChatGPT. A basic description of the database could be sent when the PM is created. Then a selection of prompts could be made available via the UI. For example, a prompt for parameterizing a query, or a prompt for adding details about a seldom used table.
My suspicion is that this could be very helpful to someone who knows a bit of SQL, but possibly it could also be implemented in a way that would help people who are new to SQL get up to speed a lot faster than they would on their own. I learned this wonderful trick today:
WITH post_type_mapping AS (
SELECT 'regular' AS type, 1 AS post_type
UNION ALL
SELECT 'moderator_action' AS type, 2 AS post_type
UNION ALL
SELECT 'small_action' AS type, 3 AS post_type
UNION ALL
SELECT 'whisper' AS type, 4 AS post_type
)
...
JOIN post_type_mapping m ON :post_type = m.type
WHERE p.post_type = m.post_type
4 Likes
Hopefully I am not feeding your addiction.
I donāt know if you read research papers but today ran across another one reinforcing that you are on a valid path and that the paper can hopefully point out new avenues for achieving the goal of generating valid SQL and results starting with a natural language query.
The paper is for math problems but since math is just an expression as is SQL, just change out one form of expression for another and it should make sense. There are many such related papers all with similar ideas so donāt consider this one as authoritative.
āSolving Challenging Math Word Problems Using Gpt-4 Code Interpreter With Code-Based Self-Verificationā by Aojun Zhoun, Ke Wang, Zimu Lu, Weikang Shi, Sichun Luo, Zipeng Qin, Shaoqing Lu, Anya Jia, Linqi Song, Mingjie Zhan and Hongsheng Li (pdf)
1 Like
Another idea that might help you and this one may seem like I am doing mushrooms but I have used it for generating Prolog code, specifically web pages.
The problem may be that generative AI has a harder time understanding SQL because it is a declarative language and much of the success with programming languages with generative AI comes from large training sets on a few imperative programming languages like JavaScript, Python, Java, etc. But generative AI which is based on transformers and transformers were initially created to translate from IIRC English to German are great at translating to/from the well trained programming languages. So if instead asking for SQL at the start instead ask for a Python or other such well trained programming language for the code to solve the problem then have the generative AI translate the Python to SQL and see if that works. I donāt plan to try this but since you seem to enjoy this, be my guest. Also if you do this, please give feedback as I would definitely like to know what you discover. 
1 Like
Is that a typo?
Should WHERE poll_id = 74 be WHERE poll_id = 71.
I did not check the entire prompt, was just looking to see if you included an example of what the results of the query should look like. In other words you gave examples of the table values but I did not see an example of the expected result for a prompt, perhaps I missed it. The expected result could be used to validate if the SQL is correct.
Suggestion:
In the example Discourse database prompt the use of 1 for id is used in multiple tables. While as humans we get that the 1 only has meaning with the context of the field and the table, the AI will not know this and that could give the AI a chance to make a choice. Therefore as a suggestion modify the Discourse database prompt to use different numbers for each id.
To go one step further, check that all of the values in each example table is a single token by using the OpenAI Tokenizer page. I know you may want to keep some of them as words or even worse strings but does the AI really care? Will values of multiple tokens lead to more variation and possible hallucinations?
1 Like
Itās a mistake. Iāll fix it. I think I re-ran the query with 71 to try to pull in results that were relevant to the other poll related tables.
I deviated from the blog postās suggestion of running all the SELECT queries with this form:
SELECT * FROM polls LIMIT 3;
I did that because Iāve got a lot of anonymized users and deleted posts in my development database. I figured Iād be supplying more coherent results this way, but Iām going to try re-doing the prompt with a fresh database and simplifying the SELECT statements.
Yeah, Iām using three users in all the examples. Their ids are 1, 2, and 121, so those values repeat a lot. I was assuming that it would be best to show coherent data. Iāll try a few different approaches and see what works best.
Another approach thatās suggested in the blog post is to limit the columns that are set in the prompt. Thatās tempting, but risks introducing a lot of errors, would be difficult to maintain, etc.
The pattern I think Iām seeing is that if I start a session by asking for a complex query, ChatGPT gets confused. I get it to work through its confusion and then the results are quite good for the rest of the session. The other approach that seems to work is to start out with a simple query and work up to more complex queries from there. Iām not sure if thatās a real pattern, or if the success rate is actually more random than it seems.
As it is now, it seems like it would be useful to someone whoās already familiar with SQL and the Discourse database. Iād like to see it get to the point where it would be useful for someone who didnāt know much about either.
Iām also testing this with ChatGPT-4. It seems likely thatās going to produce better results, but maybe be less fun to use. ChatGPT-3.5 is a lot faster.
2 Likes
Decades ago when I learned databases it was confusing just learning from the SQL but then I used the Microsoft Access query builder which was drag and drop of tables then connecting the table fields to each other with lines, much like how Visio works and that would generate the SQL.
Similar tool, image from here
Would not expect Discourse to create such a tool for building an SQL but might be possible to get the AI to generate such images of the connected tables as feedback.
IIRC in this video it is noted to first create with GPT-4 to get the correct results, then customize for GPT-3.5 for speed.
2 Likes