Structure

Layers are inferred from where files sit in the tree, not from reading the code. A project that names its directories unconventionally will read oddly here — open the file browser to check anything the diagram implies.

Languages

Share of indexed source by file size. Binary and vendored files are excluded.

Dependencies

Declared in the repository’s manifests at the indexed commit. A declared package is not proof it is used, and runtime dependencies are listed first.

Feature verification

• AI-suggested event rescheduling via TogetherAIVerified

  AtlasAI suggests event rescheduling using a fine-tuned TogetherAI model

  Claimed on Devposthigh confidence

  • src/app/recs/route.ts:7 — GET handler posts tasks/stress data to Together's inference API using a fine-tuned model and parses a rescheduled JSON event list
  • src/app/page.tsx:494 — Modal UI renders Old vs New schedule columns with reasons, sourced from recommendedTasks
• Backend built with ConvexVerified

  Backend built with Convex, integrating TerraAPI, TogetherAI and SpotifyAPI

  Claimed on Devposthigh confidence

  • convex/schema.ts:4 — defineSchema/defineTable used to declare Convex tables (tasks, recommendedTasks, stressScores, sleepScores, predictedMetrics)
  • src/app/ConvexClientProvider.tsx:1 — Convex client provider wraps the app for React Convex hooks
• Frontend built with NextJS, Tailwind, MaterialUIVerified

  Frontend built with NextJS, styling from Tailwind and MaterialUI

  Claimed on Devposthigh confidence

  • src/app/page.tsx:1 — Uses MUI components (Modal, Box, TextField, Button) and tw- prefixed Tailwind classes throughout
  • tailwind.config.ts:1 — Tailwind configuration file present
• Natural language task entry parsingVerified

  Individualized solutions using user data (task entry as part of the planner)

  Claimed on readmemedium confidence

  • src/app/page.tsx:269 — handleButtonClick parses free-text input for day, time, and duration via regex and calls sendTasks mutation to store a new task
  • convex/schedule_tasks.ts:4 — send mutation inserts the parsed task into the tasks table
• Stress score calculation and displayVerified

  AtlasAI uses individual health data (stress) to provide feedback

  Claimed on Devposthigh confidence

  • convex/stressScores.ts:20 — calculateStress normalizes HRV and heart rate into a 1-100 stress score, stored per hour
  • src/app/page.tsx:402 — Frontend queries getStressScores and displays the latest stress score and insights modal
• Wearable health data ingestion via Terra APIVerified

  Backend integrates TerraAPI to ingest wearable heart rate and sleep data

  Claimed on readmehigh confidence

  • convex/http.ts:6 — HTTP route /getjj registered as a webhook handler named getTerraAPI
  • convex/stressScores.ts:46 — getTerraAPI httpAction parses incoming heart_rate_data and computes a stress score from it
  • convex/sleepScores.ts:21 — calculateSleepData computes a sleep score from heart rate, respiration, and sleep duration fields matching Terra's data shape
• Predicted stress/sleep metricsCode-supported

  predictedMetrics table implies forward-looking health predictions

  Claimed on readmelow confidence

  • convex/predictedMetrics.ts:4 — getPredictedMetrics query reads a predictedMetrics table, but no mutation/writer populating this table or frontend consumer of it was found
  • convex/schema.ts:38 — predictedMetrics table schema defined with predictedStressScores/predictedSleepScores arrays
• Sleep score calculationCode-supported

  Health data includes sleep metrics used for recommendations

  Claimed on Devpostmedium confidence

  • convex/sleepScores.ts:21 — calculateSleepData and addSleepScores exist and are wired to the Terra webhook, but no frontend component was found reading getSleepScores
• Spotify API integrationCode-supported

  Backend integrates SpotifyAPI

  Claimed on Devpostlow confidence

  • src/app/components/SpotifyEmbedded.tsx:8 — Only a static Spotify embed iframe by playlist ID is used; no calls to the Spotify Web API (auth, search, playlist creation) were found anywhere in the repo
• Spotify playlist recommendation based on moodCode-supported

  AtlasAI recommends Spotify playlists depending on user mood

  Claimed on Devpostlow confidence

  • src/app/components/SpotifyDrawer.tsx:12 — SpotifyDrawer accepts a stress prop and embeds a Spotify playlist player, but the playlistId (6lHIfp4RnvAoXxSeLzIY49) is hardcoded rather than selected based on the stress value
  • src/app/page.tsx:907 — SpotifyDrawer is invoked with a static stress={0}, confirming mood is not actually passed through from live data
• Targeted, actionable feedback / insightsCode-supported

  AtlasAI provides targeted, actionable feedback based on health data

  Claimed on Devpostlow confidence

  • src/app/page.tsx:63 — Insights modal shows a hardcoded array of generic stress/sleep advice strings rather than data-derived per-user feedback; reason field per-task from the AI model is the more dynamic part of this claim
  • src/app/recs/route.ts:26 — Prompt requests a per-task 'reason' explaining why a task was rescheduled, tying feedback to actual stress data
• Two-phase TogetherAI fine-tuning pipelineCode-supported

  Used TogetherAI base models to generate test data, then fine-tuned on curated examples

  Claimed on Devpostmedium confidence

  • together/finetune.py:20 — Script uploads a JSONL training file and calls together.Finetune.create, supporting the fine-tuning claim, but no code proves the two-phase data-generation-then-curation workflow described
  • together/data.py:1 — Present in repo suggesting training data generation, part of the claimed fine-tuning pipeline

An AI agent derived these features from the project’s Devpost page and readme, then searched the code for each one. Verified features are backed by cited code; claimed-only features had no supporting code, which is not by itself proof a feature is missing.