Skip to content

NOTEBOOKLM.md

Latest commit

 

History

History
262 lines (207 loc) · 7.29 KB

NOTEBOOKLM.md

File metadata and controls

262 lines (207 loc) · 7.29 KB

DevAIFlow - AI-Optimized Summary

Comprehensive overview for AI tools (NotebookLM, Claude, etc.) to understand DevAIFlow capabilities.

What is DevAIFlow?

Session management tool for AI coding assistants with optional JIRA/GitHub/GitLab integration. Organizes development work, tracks time, and coordinates multi-repository projects.

Core Value: Bridges issue trackers and AI-assisted development with context management, time tracking, and workflow automation.

Key Capabilities

  1. Session Management - Organize AI sessions with goals, time tracking, progress notes
  2. Multi-AI Assistant Support - Claude Code (fully tested), GitHub Copilot, Cursor, Windsurf (experimental)
  3. Alternative Model Providers - Run with local models (llama.cpp) or cloud providers (OpenRouter, Vertex AI) - save up to 98% on costs or run offline
  4. Issue Tracker Integration - JIRA, GitHub Issues, GitLab Issues (or none)
  5. Multi-Repository Support - Work across repos with shared AI context
  6. Automatic Time Tracking - Per session/ticket tracking
  7. Team Collaboration - Export/import sessions for handoffs
  8. Workflow Automation - Auto commits, PR/MR creation, ticket transitions

Core Concepts

Session: Focused work on one task (one JIRA ticket/GitHub issue)

  • Unique name/ID, time tracking, progress notes
  • Can span multiple repositories (multi-project)
  • States: created → in_progress → complete

Conversation: Claude Code conversation within a session

  • Multi-project: ONE conversation with SHARED CONTEXT across all projects
  • Legacy: Separate conversations per repository
  • Each has own git branch

Workspace: Named directory for repositories (enables concurrent multi-branch development)

Three Essential Workflows

1. Create Tickets with Analysis (New Work)

JIRA:

daf jira new story --parent PROJ-123 --goal "Add feature X"
# Claude analyzes codebase (read-only), creates informed ticket
daf complete <session>
daf open PROJ-12345

GitHub/GitLab:

daf git new --goal "Add feature X"
# Claude analyzes, creates issue
daf complete <session>
daf open owner-repo-123

2. Sync Existing Tickets (Assigned Work)

JIRA:

daf sync --sprint current
daf open PROJ-12345
daf complete PROJ-12345

GitHub/GitLab:

daf sync
daf open owner-repo-123
daf complete owner-repo-123

3. Multi-Project Development

Declarative:

daf new PROJ-123 --projects backend-api,frontend-app,shared-lib
# ONE conversation with SHARED CONTEXT across all 3 projects
daf complete PROJ-123  # Creates 3 PRs

Iterative:

daf open PROJ-123    # Select backend-api
daf open PROJ-123    # Add frontend-app
daf complete PROJ-123  # Creates 2 PRs

Essential Commands

Session

daf new --name "..." --goal "..."
daf open <session>
daf complete <session>
daf list / daf list --active
daf status

Ticket Creation

daf jira new <type> --parent <key> --goal "..."
daf git new --goal "..."
daf jira create <type> --summary "..." --parent <key>
daf git create --summary "..."

Sync

daf sync
daf sync --workspace <name>
daf sync --sprint current
daf sync --type Story

Progress

daf note <session> "Update"
daf notes <session>
daf summary <session>
daf time <session>

Issue Tracker

daf jira view <key>
daf jira add-comment <key> "..."
daf git view "owner/repo#123"
daf git add-comment "owner/repo#123" "..."

Maintenance

daf maintenance cleanup-conversation <session> --older-than 8h
daf export <session> --output file.tar.gz
daf import file.tar.gz
daf template save <session> <name>

Architecture

Single-Session:

Session: PROJ-12345
└── Conversation (backend-api)
    ├── Git branch
    ├── Claude session
    └── Time tracking

Multi-Project:

Session: PROJ-12345
└── Conversation (SHARED CONTEXT)
    ├── backend-api (branch)
    ├── frontend-app (branch)
    ├── shared-lib (branch)
    └── Unified time tracking

Issue Tracker Support

Backend CLI Features
JIRA JIRA CLI/API Workflow automation, custom fields, sprints
GitHub gh Issue sync, PR creation, comments
GitLab glab Issue sync, MR creation, comments
None - Local session management

Key Features

Codebase Analysis Before Tickets

Claude analyzes code BEFORE creating tickets → better descriptions and acceptance criteria

Multi-Project Shared Context

ONE conversation across all repos → Claude coordinates changes (not separate conversations)

Concurrent Development

Different workspaces = same project, different branches, no conflicts

Session Templates

daf template save PROJ-12345 backend-api
daf new --template backend-api --goal "..."

Team Handoffs

daf export PROJ-12345 --output handoff.tar.gz
daf import handoff.tar.gz

Use Cases

  1. Sprint work with JIRA (sync tickets, track time, auto PR/status)
  2. Open source GitHub (sync issues, multi-repo changes, create PRs)
  3. Multi-repo features (backend + frontend in one session)
  4. Personal projects (no issue tracker, just sessions)
  5. Team handoffs (export/import with full context)
  6. Code investigations (analyze before creating tickets)

Comparison: Traditional vs DevAIFlow

Traditional:

  1. Create ticket (no code context)
  2. Open IDE, read code
  3. Code
  4. Manual time tracking
  5. Create PR
  6. Update ticket

DevAIFlow:

  1. daf jira new → Claude analyzes, creates informed ticket
  2. daf open → Start with full context
  3. Auto time tracking
  4. daf complete → Commits, PR, ticket update auto

Important Notes

Session Naming:

  • JIRA: PROJ-12345
  • GitHub session: owner-repo-123 (no quotes)
  • GitHub issue key: "owner/repo#123" (quotes required, # starts comments in bash)

Multi-Project:

  • Declarative (--projects): Faster, all at once
  • Iterative (multiple opens): Flexible, discover as you go

Configuration:

$DEVAIFLOW_HOME/
├── config.json
├── backends/jira.json
├── organization.json
├── team.json
└── sessions/<session-name>/

Alternative Model Providers

DevAIFlow supports running Claude Code with alternative AI models for cost savings or offline use:

Local Models (100% offline):

  • llama.cpp - Run models completely locally, no internet required
  • LM Studio - Alternative local model runner

Cloud Providers (98% cheaper):

  • OpenRouter - Access multiple models at lower cost
  • Vertex AI - Google Cloud AI models
  • Minimax - Chinese market AI provider

Configuration: Use daf config edit to configure model provider profiles. See docs/alternative-model-providers.md for setup instructions.

⚠️ Note: Ollama is NOT compatible with Claude Code. Use llama.cpp instead.

Quick Links