* FEATURE: Native PDF support
This amends it so we use PDF Reader gem to extract text from PDFs
* This means that our simple pdf eval passes at last
* fix spec
* skip test in CI
* test file support
* Update lib/utils/image_to_text.rb
Co-authored-by: Alan Guo Xiang Tan <gxtan1990@gmail.com>
* address pr comments
---------
Co-authored-by: Alan Guo Xiang Tan <gxtan1990@gmail.com>
This PR introduces several enhancements and refactorings to the AI Persona and RAG (Retrieval-Augmented Generation) functionalities within the discourse-ai plugin. Here's a breakdown of the changes:
**1. LLM Model Association for RAG and Personas:**
- **New Database Columns:** Adds `rag_llm_model_id` to both `ai_personas` and `ai_tools` tables. This allows specifying a dedicated LLM for RAG indexing, separate from the persona's primary LLM. Adds `default_llm_id` and `question_consolidator_llm_id` to `ai_personas`.
- **Migration:** Includes a migration (`20250210032345_migrate_persona_to_llm_model_id.rb`) to populate the new `default_llm_id` and `question_consolidator_llm_id` columns in `ai_personas` based on the existing `default_llm` and `question_consolidator_llm` string columns, and a post migration to remove the latter.
- **Model Changes:** The `AiPersona` and `AiTool` models now `belong_to` an `LlmModel` via `rag_llm_model_id`. The `LlmModel.proxy` method now accepts an `LlmModel` instance instead of just an identifier. `AiPersona` now has `default_llm_id` and `question_consolidator_llm_id` attributes.
- **UI Updates:** The AI Persona and AI Tool editors in the admin panel now allow selecting an LLM for RAG indexing (if PDF/image support is enabled). The RAG options component displays an LLM selector.
- **Serialization:** The serializers (`AiCustomToolSerializer`, `AiCustomToolListSerializer`, `LocalizedAiPersonaSerializer`) have been updated to include the new `rag_llm_model_id`, `default_llm_id` and `question_consolidator_llm_id` attributes.
**2. PDF and Image Support for RAG:**
- **Site Setting:** Introduces a new hidden site setting, `ai_rag_pdf_images_enabled`, to control whether PDF and image files can be indexed for RAG. This defaults to `false`.
- **File Upload Validation:** The `RagDocumentFragmentsController` now checks the `ai_rag_pdf_images_enabled` setting and allows PDF, PNG, JPG, and JPEG files if enabled. Error handling is included for cases where PDF/image indexing is attempted with the setting disabled.
- **PDF Processing:** Adds a new utility class, `DiscourseAi::Utils::PdfToImages`, which uses ImageMagick (`magick`) to convert PDF pages into individual PNG images. A maximum PDF size and conversion timeout are enforced.
- **Image Processing:** A new utility class, `DiscourseAi::Utils::ImageToText`, is included to handle OCR for the images and PDFs.
- **RAG Digestion Job:** The `DigestRagUpload` job now handles PDF and image uploads. It uses `PdfToImages` and `ImageToText` to extract text and create document fragments.
- **UI Updates:** The RAG uploader component now accepts PDF and image file types if `ai_rag_pdf_images_enabled` is true. The UI text is adjusted to indicate supported file types.
**3. Refactoring and Improvements:**
- **LLM Enumeration:** The `DiscourseAi::Configuration::LlmEnumerator` now provides a `values_for_serialization` method, which returns a simplified array of LLM data (id, name, vision_enabled) suitable for use in serializers. This avoids exposing unnecessary details to the frontend.
- **AI Helper:** The `AiHelper::Assistant` now takes optional `helper_llm` and `image_caption_llm` parameters in its constructor, allowing for greater flexibility.
- **Bot and Persona Updates:** Several updates were made across the codebase, changing the string based association to a LLM to the new model based.
- **Audit Logs:** The `DiscourseAi::Completions::Endpoints::Base` now formats raw request payloads as pretty JSON for easier auditing.
- **Eval Script:** An evaluation script is included.
**4. Testing:**
- The PR introduces a new eval system for LLMs, this allows us to test how functionality works across various LLM providers. This lives in `/evals`
When you already have embeddings for a model stored and change models,
our backfill script was failing to backfill the newly configured model.
Regression introduced most likely in 1686a8a
Before this change, a summary was only outdated when new content appeared, for topics with "best replies", when the query returned different results. The intent behind this change is to detect when a summary is outdated as a result of an edit.
Additionally, we are changing the backfill candidates query to compare "ai_summary_backfill_topic_max_age_days" against "last_posted_at" instead of "created_at", to catch long-lived, active topics. This was discussed here: https://meta.discourse.org/t/ai-summarization-backfill-is-stuck-keeps-regenerating-the-same-topic/347088/14?u=roman_rizzi
* Use AR model for embeddings features
* endpoints
* Embeddings CRUD UI
* Add presets. Hide a couple more settings
* system specs
* Seed embedding definition from old settings
* Generate search bit index on the fly. cleanup orphaned data
* support for seeded models
* Fix run test for new embedding
* fix selected model not set correctly
To quickly select backfill candidates without comparing SHAs, we compare the last summarized post to the topic's highest_post_number. However, hiding or deleting a post and adding a small action will update this column, causing the job to stall and re-generate the same summary repeatedly until someone posts a regular reply. On top of this, this is not always true for topics with `best_replies`, as this last reply isn't necessarily included.
Since this is not evident at first glance and each summarization strategy picks its targets differently, I'm opting to simplify the backfill logic and how we track potential candidates.
The first step is dropping `content_range`, which serves no purpose and it's there because summary caching was supposed to work differently at the beginning. So instead, I'm replacing it with a column called `highest_target_number`, which tracks `highest_post_number` for topics and could track other things like channel's `message_count` in the future.
Now that we have this column when selecting every potential backfill candidate, we'll check if the summary is truly outdated by comparing the SHAs, and if it's not, we just update the column and move on
In a previous refactor, we moved the responsibility of querying and storing embeddings into the `Schema` class. Now, it's time for embedding generation.
The motivation behind these changes is to isolate vector characteristics in simple objects to later replace them with a DB-backed version, similar to what we did with LLM configs.
* FIX: Make sure gists are atleast five minutes old before updating them
* Update app/jobs/regular/fast_track_topic_gist.rb
Co-authored-by: Keegan George <kgeorge13@gmail.com>
---------
Co-authored-by: Keegan George <kgeorge13@gmail.com>
* REFACTOR: A Simpler way of interacting with embeddings' tables.
This change adds a new abstraction called `Schema`, which acts as a repository that supports the same DB features `VectorRepresentation::Base` has, with the exception that removes the need to have duplicated methods per embeddings table.
It is also a bit more flexible when performing a similarity search because you can pass it a block that gives you access to the builder, allowing you to add multiple joins/where conditions.
This introduces a comprehensive spam detection system that uses LLM models
to automatically identify and flag potential spam posts. The system is
designed to be both powerful and configurable while preventing false positives.
Key Features:
* Automatically scans first 3 posts from new users (TL0/TL1)
* Creates dedicated AI flagging user to distinguish from system flags
* Tracks false positives/negatives for quality monitoring
* Supports custom instructions to fine-tune detection
* Includes test interface for trying detection on any post
Technical Implementation:
* New database tables:
- ai_spam_logs: Stores scan history and results
- ai_moderation_settings: Stores LLM config and custom instructions
* Rate limiting and safeguards:
- Minimum 10-minute delay between rescans
- Only scans significant edits (>10 char difference)
- Maximum 3 scans per post
- 24-hour maximum age for scannable posts
* Admin UI features:
- Real-time testing capabilities
- 7-day statistics dashboard
- Configurable LLM model selection
- Custom instruction support
Security and Performance:
* Respects trust levels - only scans TL0/TL1 users
* Skips private messages entirely
* Stops scanning users after 3 successful public posts
* Includes comprehensive test coverage
* Maintains audit log of all scan attempts
---------
Co-authored-by: Keegan George <kgeorge13@gmail.com>
Co-authored-by: Martin Brennan <martin@discourse.org>
* FEATURE: Backfill posts sentiment.
It adds a scheduled job to backfill posts' sentiment, similar to our existing rake task, but with two settings to control the batch size and posts' max-age.
* Make sure model_name order is consistent.
This change adds a simpler class for sentiment classification, replacing the soon-to-be removed `Classificator` hierarchy. Additionally, it adds a method for classifying concurrently, speeding up the backfill rake task.
We are adding a new method for generating and storing embeddings in bulk, which relies on `Concurrent::Promises::Future`. Generating an embedding consists of three steps:
Prepare text
HTTP call to retrieve the vector
Save to DB.
Each one is independently executed on whatever thread the pool gives us.
We are bringing a custom thread pool instead of the global executor since we want control over how many threads we spawn to limit concurrency. We also avoid firing thousands of HTTP requests when working with large batches.
This change introduces a job to summarize topics and cache the results automatically. We provide a setting to control how many topics we'll backfill per hour and what the topic's minimum word count is to qualify.
We'll prioritize topics without summary over outdated ones.
* FEATURE: Fast-track gist regeneration when a hot topic gets a new post
* DEV: Introduce an upsert-like summarize
* FIX: Only enqueue fast-track gist for hot hot hot topics
---------
Co-authored-by: Rafael Silva <xfalcox@gmail.com>
* Display gists in the hot topics list
* Adjust hot topics gist strategy and add a job to generate gists
* Replace setting with a configurable batch size
* Avoid loading summaries for other topic lists
* Tweak gist prompt to focus on latest posts in the context of the OP
* Remove serializer hack and rely on core change from discourse/discourse#29291
* Update lib/summarization/strategies/hot_topic_gists.rb
Co-authored-by: Rafael dos Santos Silva <xfalcox@gmail.com>
---------
Co-authored-by: Rafael dos Santos Silva <xfalcox@gmail.com>
On very large sites, the rare cache misses for Related Topics can take around 200ms, which affects our p99 metric on the topic page. In order to mitigate this impact, we now have several tools at our disposal.
First, one is to migrate the index embedding type from halfvec to bit and change the related topic query to leverage the new bit index by changing the search algorithm from inner product to Hamming distance. This will reduce our index sizes by 90%, severely reducing the impact of embeddings on our storage. By making the related query a bit smarter, we can have zero impact on recall by using the index to over-capture N*2 results, then re-ordering those N*2 using the full halfvec vectors and taking the top N. The expected impact is to go from 200ms to <20ms for cache misses and from a 2.5GB index to a 250MB index on a large site.
Another tool is migrating our index type from IVFFLAT to HNSW, which can increase the cache misses performance even further, eventually putting us in the under 5ms territory.
Co-authored-by: Roman Rizzi <roman@discourse.org>
This allows custom tools access to uploads and sophisticated searches using embedding.
It introduces:
- A shared front end for listing and uploading files (shared with personas)
- Backend implementation of index.search function within a custom tool.
Custom tools now may search through uploaded files
function invoke(params) {
return index.search(params.query)
}
This means that RAG implementers now may preload tools with knowledge and have high fidelity over
the search.
The search function support
specifying max results
specifying a subset of files to search (from uploads)
Also
- Improved documentation for tools (when creating a tool a preamble explains all the functionality)
- uploads were a bit finicky, fixed an edge case where the UI would not show them as updated
Polymorphic RAG means that we will be able to access RAG fragments both from AiPersona and AiCustomTool
In turn this gives us support for richer RAG implementations.
Previously we waited 1 minute before automatically titling PMs
The new change introduces adding a title immediately after the the
llm replies
Prompt was also modified to include the LLM reply in title suggestion.
This helps situation like:
user: tell me a joke
llm: a very funy joke about horses
Then the title would be "A Funny Horse Joke"
Specs already covered some auto title logic, amended to also
catch the new message bus message we have been sending.
* FIX: we were never reindexing old content
Embedding backfill contains logic for searching for old content
change and then backfilling.
Unfortunately it was excluding all topics that had embedding
unconditionally, leading to no backfill ever happening.
This change adds a test and ensures we backfill.
* over select results, this ensures we will be more likely to find
ai results when filtered
When navigating between topic we were not correctly resetting
internal state for summarization. This leads to a situation where
incorrect summaries can be displayed to users and wrong summaries
can be displayed.
Additionally our controller for grabbing summaries was always
streaming results via message bus, which could be delayed when
sidekiq is overloaded. We now will return the cached summary
right away if it is available direct from REST endpoint.
This allows summary to use the new LLM models and migrates of API key based model selection
Claude 3.5 etc... all work now.
---------
Co-authored-by: Roman Rizzi <rizziromanalejandro@gmail.com>
Add support for chat with AI personas
- Allow enabling chat for AI personas that have an associated user
- Add new setting `allow_chat` to AI persona to enable/disable chat
- When a message is created in a DM channel with an allowed AI persona user, schedule a reply job
- AI replies to chat messages using the persona's `max_context_posts` setting to determine context
- Store tool calls and custom prompts used to generate a chat reply on the `ChatMessageCustomPrompt` table
- Add tests for AI chat replies with tools and context
At the moment unlike posts we do not carry tool calls in the context.
No @mention support yet for ai personas in channels, this is future work
Updating the editing model's rag_uploads in the editor component broke multi-file uploading. Instead, we'll keep the uploads in the uploader and update the model when we finish.
This PR also fast-tracks the initial update so we can show feedback to the user quickly, and allows uploading MD files.
Bug reported on https://meta.discourse.org/t/discourse-ai-persona-upload-support/304049/11
* FEATURE: allow tuning of RAG generation
- change chunking to be token based vs char based (which is more accurate)
- allow control over overlap / tokens per chunk and conversation snippets inserted
- UI to control new settings
* improve ui a bit
* fix various reindex issues
* reduce concurrency
* try ultra low queue ... concurrency 1 is too slow.
This commit uses a new plugin modifier introduced in https://github.com/discourse/discourse/pull/26508
to mark all uploads as _not_ secure in shared PM AI conversations.
This is so images created by the AI bot (or uploaded by the user)
do not end up as broken URLs because of the security requirements
around them.
This relies on the UpdateTopicUploadSecurity job in core as well,
which is fired when an AI conversation is shared or deleted.
* FEATURE: Add metadata support for RAG
You may include non indexed metadata in the RAG document by using
[[metadata ....]]
This information is attached to all the text below and provided to
the retriever.
This allows for RAG to operate within a rich amount of contexts
without getting lost
Also:
- re-implemented chunking algorithm so it streams
- moved indexing to background low priority queue
* Baran gem no longer required.
* tokenizers is on 4.4 ... upgrade it ...
This PR lets you associate uploads to an AI persona, which we'll split and generate embeddings from. When building the system prompt to get a bot reply, we'll do a similarity search followed by a re-ranking (if available). This will let us find the most relevant fragments from the body of knowledge you associated with the persona, resulting in better, more informed responses.
For now, we'll only allow plain-text files, but this will change in the future.
Commits:
* FEATURE: RAG embeddings for the AI Bot
This first commit introduces a UI where admins can upload text files, which we'll store, split into fragments,
and generate embeddings of. In a next commit, we'll use those to give the bot additional information during
conversations.
* Basic asymmetric similarity search to provide guidance in system prompt
* Fix tests and lint
* Apply reranker to fragments
* Uploads filter, css adjustments and file validations
* Add placeholder for rag fragments
* Update annotations
