OpenAI Integration — Fine-tuning data sources and eligibility
Prepare eligible public knowledge-base articles and predefined replies and understand the minimum-data checks.
01Open the OpenAI fine-tuning workspaceAccess training-source counts, job status and model controls from the OpenAI settings page.Admin Area → Setup → Settings → AI → OpenAI → Advanced Features → OpenAI Fine Tuning02Understand fine-tuning permissionsSeparate read access from job, model and option changes.Admin Area → Setup → Settings → AI → OpenAI → Advanced Features → OpenAI Fine Tuning03Choose a fine-tuning base modelSelect the supported base model used for the next fine-tuning job.Admin Area → Setup → Settings → AI → OpenAI → Advanced Features → OpenAI Fine Tuning04Prepare eligible knowledge-base articles for fine-tuningPublish active public articles that the training collector can read.Admin Area → Support/Knowledge Base → Articles, then OpenAI Fine Tuning05Prepare predefined replies for fine-tuningCreate approved reusable replies that the training collector can include.Admin Area → Setup/Support → Predefined Replies, then OpenAI Fine Tuning06Meet the minimum fine-tuning source requirementProvide enough eligible content for the Start Fine Tuning action.Admin Area → OpenAI Fine Tuning → Training Data Sources07Review knowledge-base and predefined-reply source countsUse the displayed counts before starting a job.Admin Area → Setup → Settings → AI → OpenAI → Advanced Features → OpenAI Fine Tuning08Understand the default fine-tuning training promptKnow how each source item is converted into a user/assistant example.Code-backed reference → OpenAiProvider::formatTrainingData09Extend eligible knowledge-base sources with the openai_fine_tuning_knowledge_base filterAdjust the collected public article array through the supported hook.Development integration → hook openai_fine_tuning_knowledge_base10Extend predefined-reply sources with the openai_fine_tuning_predefined_replies filterAdjust the predefined-reply training array through the supported hook.Development integration → hook openai_fine_tuning_predefined_replies11Alter formatted training examples with the openai_fine_tuning_training_data filterModify the final message arrays before JSONL encoding.Development integration → hook openai_fine_tuning_training_data12Troubleshoot insufficient fine-tuning dataResolve the start-job rejection when neither source reaches the required count.Admin Area → Setup → Settings → AI → OpenAI → Advanced Features → OpenAI Fine Tuning