Skip to content

ROADMAP.md

Latest commit

 

History

History
225 lines (190 loc) · 10.2 KB

ROADMAP.md

File metadata and controls

225 lines (190 loc) · 10.2 KB

Riona AI Agent Roadmap

This roadmap is based on the current codebase state and is meant to guide ongoing development. Items marked as completed are already implemented in the repository.

Status Legend

Done In progress Planned Blocked

Summary

  • Current state: Instagram automation, AI content generation, training utilities, API server, dashboard, and reCAPTCHA ML subproject are present.
  • Near term: Stabilize automation, finish Twitter/X pipeline, improve safety/rate limits, and harden ops.
  • Long term: Multi‑platform expansion, richer training, analytics, and compliance tooling.

Milestones

  • Core API server, auth, and health endpoints
  • Instagram automation (login, interact loop, posting, scheduling)
  • AI comment generation with Gemini + schema
  • Training utilities (YouTube, file parsing, website scraping)
  • Simple dashboard for health + last run
  • reCAPTCHA ML service (train/serve/admin UI)
  • Twitter/X full workflow (compose, schedule, media, metrics)
  • GitHub automation (planned)
  • Analytics and reporting (cross‑platform)
  • Production hardening and compliance features

Phased Delivery

| Phase | Goals | Status | | --- | --- | --- | --- | | Phase 1 | Twitter/X MVP, IG reliability pass, admin UI basics | In progress | | Phase 2 | Analytics, observability, policy/rules engine, compliance guardrails | Planned | | Phase 3 | Multi‑platform orchestration, model evaluation + A/B, scale & cost controls | Planned |

Workstreams and Checklist

1) Platform Automation

  • Instagram browser automation (Puppeteer + stealth)
  • Instagram post by URL and file
  • Instagram scheduling (cron)
  • Instagram follower scraping
  • Cookie persistence and relogin handling
  • Instagram reliability pass: captcha/challenge escalation workflow
  • Instagram action throttling by risk profile (dynamic)
  • Twitter/X end‑to‑end publish pipeline
  • Twitter/X scheduling and media upload
  • Twitter/X engagement actions (like/retweet/reply)
  • GitHub automation: issues, PRs, releases

2) AI and Training

  • Gemini JSON‑schema generation for comments
  • API key rotation on rate limit
  • YouTube transcript ingestion
  • Audio and file‑based training ingestion
  • Website scraping for training data
  • Prompt evaluation harness with golden datasets
  • Multi‑persona training and selection
  • Safety filters for toxic content and brand rules
  • Model/agent selection and A/B testing

3) Data and Storage

  • MongoDB connection and models
  • Tweet schema for rate limiting
  • Unified action log (IG/Twitter/GitHub)
  • Content cache + dedupe layer
  • Audit trail for moderation and compliance

4) API and Dashboard

  • REST API (login, interact, post, schedule)
  • /dashboard summary UI
  • Health endpoint
  • Web UI for configuring accounts and profiles
  • Admin panel for viewing actions, logs, and errors
  • Webhook endpoints for external triggers
  • API rate limiting and API keys for third‑party usage

5) Ops, Security, Reliability

  • Env validation scripts
  • Logging with Winston
  • Docker‑based MongoDB setup docs
  • Secrets management for production (vault/parameter store)
  • Observability: metrics + alerts
  • CI coverage for integration tests
  • Chaos testing for IG loops
  • Legal/compliance guardrails and TOS risk toggles
  • Action audit logs and tamper‑evident storage
  • Automated rollback on error spikes
  • Rate‑limit aware backoff across providers

6) reCAPTCHA ML Subproject

  • Model architecture and training flow
  • Admin UI and debug views
  • Training data collection pipeline
  • Model serving endpoints
  • Dataset versioning and quality checks
  • Model performance tracking and drift detection
  • Active learning loop for hard examples
  • Eval harness with fixed validation sets

Timeline (High‑Level)

%%{init: {'theme': 'base', 'themeVariables': {
  'primaryColor': '#e9f1ff',
  'primaryTextColor': '#1d4ed8',
  'primaryBorderColor': '#c7dbff',
  'secondaryColor': '#e7f8ef',
  'secondaryTextColor': '#0f7b47',
  'secondaryBorderColor': '#bfead4',
  'tertiaryColor': '#fff4e5',
  'tertiaryTextColor': '#8a4f00',
  'tertiaryBorderColor': '#ffd9a8'
}}}%%
timeline
  title Riona AI Agent Roadmap
  Current (Done) : Core API + auth
                 : Instagram automation
                 : AI comment generation
                 : Training utilities
                 : Dashboard + health
                 : reCAPTCHA ML service
  Next (Planned) : Twitter/X full workflow
                 : Reliability + rate limiting
                 : Admin UI + config flows
                 : Analytics + reporting
  Later (Planned) : GitHub automation
                  : Compliance + safety controls
                  : Model evaluation + A/B testing
                  : Multi‑platform orchestration
Loading

Architecture Coverage (Now vs Target)

%%{init: {'theme': 'base', 'themeVariables': {
  'fontFamily': 'Inter, Segoe UI, Arial',
  'primaryColor': '#e9f1ff',
  'primaryTextColor': '#1d4ed8',
  'primaryBorderColor': '#c7dbff',
  'lineColor': '#94a3b8'
}}}%%
flowchart LR
  subgraph Now
    A1[API Server] --> A2[Instagram Bot]
    A1 --> A3[Gemini Agent]
    A1 --> A4[MongoDB]
    A1 --> A5[Dashboard]
    A6[reCAPTCHA Service] --> A7[TFJS Model]
  end

  subgraph Target
    B1[API Gateway] --> B2[IG Bot]
    B1 --> B3[X/Twitter Bot]
    B1 --> B4[GitHub Bot]
    B1 --> B5[Analytics]
    B1 --> B6[Admin UI]
    B1 --> B7[Compliance]
    B8[Training Hub] --> B9[Eval + A/B]
  end

  classDef done fill:#e7f8ef,stroke:#0f7b47,color:#0f7b47;
  classDef active fill:#fff4e5,stroke:#8a4f00,color:#8a4f00;
  classDef planned fill:#e9f1ff,stroke:#1d4ed8,color:#1d4ed8;
  classDef target fill:#f6f7fb,stroke:#64748b,color:#334155;

  class A1,A2,A3,A4,A5,A6,A7 done;
  class B1,B2,B3,B4,B5,B6,B7,B8,B9 planned;
Loading

Delivery Criteria (Release Gates)

Gate Criteria Status
Reliability 7‑day crash‑free IG loop, < 3% challenge rate Planned
Security Secrets not stored in repo, JWT/session hardened Planned
Quality Integration tests for login/post/cron Planned
Observability Metrics + alerts for errors and cooldowns Planned

KPI Targets (Suggested)

Area KPI Baseline Target Notes
Instagram Successful interactions per run TBD +30% Based on stable login + cooldowns
Instagram Challenge rate TBD < 3% Requires risk‑based throttling
AI Output Avg comment engagement TBD +25% Needs analytics + tracking
Reliability Crash‑free loop runs TBD 99% Add watchdog + retries
ML Model reCAPTCHA accuracy TBD > 92% Track via validation set

Risks and Mitigations

Risk Impact Mitigation Status
IG challenges / bans High Dynamic throttling, cooldowns, better fingerprinting In progress
Provider rate limits Medium Key rotation, backoff, queueing Done
Data quality drift Medium Validation sets, drift monitoring Planned
Compliance exposure High Policy rules, opt‑out, audit logs Planned

Dependencies

  • Stable access to IG and X/Twitter APIs or browser‑automation compatibility.
  • Gemini API key availability and quota.
  • MongoDB availability for persistence and logging.
  • Proxy infrastructure if scaling across multiple accounts.

Quality and Testing Strategy

  • Unit tests for utilities and schema constraints.
  • Integration tests for login, posting, scheduling, and cooldown behavior.
  • Load tests for API endpoints and training pipelines.
  • Regression suite for prompt outputs and content safety filters.

Backlog (Unscheduled)

  • Mobile‑friendly dashboard
  • Multi‑account policy/rules engine
  • Task queue (BullMQ/Redis)
  • Content calendar and approvals
  • Per‑account proxy assignment
  • Pluggable model providers (OpenAI, local)
  • Data retention and deletion policies
  • Webhook‑based partner integrations