Case study
Social Media Assistant
A guided content workflow for a fitness trainer maintaining an Instagram blog with reviewed health and fitness source material.
The assistant helps a fitness trainer maintain an Instagram blog while preserving their editorial voice. It turns health and fitness source material into reviewed content that the trainer can preview, adjust, and control.
The product is organised around a creator profile and a deliberate editorial pipeline: useful sources are collected, assessed, and shaped into a post through clear review gates.
It brings scattered editorial decisions into one controlled content workflow while preserving the trainer's judgement.
A profile before a publishing schedule
The experience starts with Telegram guided onboarding rather than a blank request for a post. The trainer can define content directions, maintain editable topics, select a tone and platform, and choose delivery preferences that make the workflow fit their Instagram blog.
Those choices remain editable as the blog evolves. Feedback and style adjustments are part of the product, so a creator can refine the profile instead of treating the initial configuration as a permanent instruction.
This gives the system a usable editorial frame while keeping the trainer in control of the subject matter, language, and final content choice.
From source material to an editorial candidate
The assistant ingests material from trusted RSS sources and PubMed to supply potential health and fitness topics for review.
Deterministic eligibility and quality checks create the first editorial gate, focusing generative work on material that meets the product's standards. AI then supports source selection, post generation, and quality review in that order.
Prepared content moves through one shared manual and scheduled delivery pipeline. Telegram provides preview and control for feedback and manual delivery, while scheduled work supports seamless Instagram autopost.
A reviewable content path
The path keeps source evidence, editorial gates, creator review, and Instagram delivery visible as separate decisions.
Step 1
Trainer profile
Set directions, editable topics, tone, platform, and delivery preferences for the fitness trainer's Instagram blog.
Step 2
Trusted RSS collection
Collect potential health and fitness material from trusted RSS sources.
Step 3
PubMed research
Bring PubMed material into the same source-review path alongside trusted RSS inputs.
Step 4
Deterministic eligibility and quality
Apply deterministic eligibility and quality checks before generative editorial work.
Step 5
AI source selection
Use AI source selection to identify a suitable candidate from material that passed the deterministic gates.
Step 6
Post generation
Prepare a post generation candidate in the voice and topic boundaries defined by the trainer profile.
Step 7
Quality review
Run AI quality review and preserve a clear point for creator judgement before delivery.
Step 8
Telegram preview and control
Use Telegram preview and control for feedback and the manual delivery path when the trainer wants to intervene.
Step 9
Seamless Instagram autopost
Deliver scheduled content to the trainer's Instagram blog as part of the editorial workflow.
The services behind a controlled workflow
The product is a TypeScript Node backend built with Telegraf and Supabase repositories and migrations. Its modules separate the bot, source collection, generation, delivery, feedback, and style responsibilities, keeping editorial actions clear and traceable.
OpenAI role clients support the AI stages, while a scheduler supports planned delivery. Telegram is the control surface for preview, review, feedback, and manual choices; Supabase persists the profile, source, content, and delivery state that keeps those decisions traceable.
Instagram Graph API integration brings publication into the same editorial workflow, linking source review, AI post generation, and delivery for the trainer's Instagram blog.