v0.10.2 — PyPI README Images & Presentation Fix

Agentic Career Lab v0.10.2

Agentic Career Lab v0.10.2 updates documentation assets and URLs for full compatibility with PyPI and external packaging renderers.

Changes in v0.10.2

• PyPI Image Rendering: Migrated application screenshot and architecture diagram links to absolute GitHub URLs (raw.githubusercontent.com) so they render seamlessly on both GitHub and PyPI package pages.
• Documentation Link Portability: Updated relative documentation links to absolute URLs for consistent navigation across external registry pages.
• Verified Suite: 113 automated tests passing, clean Ruff formatting, and validated packaging metadata.

v0.10.1 — PyPI Release and UVX CLI Launcher

Agentic Career Lab v0.10.1

Agentic Career Lab v0.10.1 introduces official PyPI distribution and the agentic-career-lab CLI launcher for instant, isolated execution via uvx.

Highlights

• First PyPI Publication: Direct installation via pip install agentic-career-lab or one-liner launching via uvx agentic-career-lab.
• CLI Launcher & Doctor Diagnostics: Built-in agentic-career-lab doctor command to verify Python runtime, Ollama connectivity, Gemma 4 model availability, and network ports.
• Trusted Publishing: Automated, tokenless PyPI releases powered by GitHub Actions OpenID Connect (OIDC).
• Preserved Core Features: Retains full v0.10.0 multi-agent workflow, local private Gemma 4 LLM analysis, synthetic fallback resumes, and interactive Ask Career AI chat.

Agentic Career Lab v0.10.0 — Ask Career AI & Interactive Actions

Agentic Career Lab v0.10.0

What's New

This release introduces Ask Career AI, a privacy-first conversational career assistant powered by local Gemma 4 through Ollama.

Highlights

• Multi-turn Streamlit AI career chat
• Streaming local Gemma 4 responses
• Grounded career tools and agent coordination
• Job selection directly from chat
• Evidence-based skill gap navigation
• Curated learning resource links
• Continue Learning action
• Career Report navigation
• Privacy-conscious resume processing
• Improved model timeout handling
• Automated browser acceptance coverage

Technology

• Python
• Streamlit
• Google ADK integration
• Gemma 4
• Ollama
• Docker Compose
• OpenTelemetry
• Prometheus
• Grafana
• Tempo

Privacy

Local inference is the default.

Uploaded resumes are processed locally.
Cloud execution requires explicit consent.

Validation

• 102 pytest tests passed in pre-release testing
• 10-stage career workflow browser validation
• Five interactive career action browser flows

Known Limitations

• Chat history is session-based.
• Resume Studio editing is planned for v0.11.0.
• Persistent workspace storage is planned for v0.12.0.
• Local inference may be slower during model cold starts.

For setup and quickstart instructions, refer to the README.md.

v0.9.0 — Guided Learning Platform and Career Journey

Agentic Career Lab v0.9.0 — Guided Learning Platform and Career Journey

A privacy-first hybrid multi-agent career platform for students and technical job seekers.

🚀 Key Highlights & Verified Capabilities

• Dedicated My Profile Page: Academic background, graduation year, target career role, location preferences, and demo student profile loader.
• Dashboard Quick Actions: One-click direct exploration cards (Set Up My Profile, Find Matching Jobs, Compare My Resume, Identify Skill Gaps).
• Interactive Career Journey Tracker: Clickable, persistent 3-stage workflow tracker across all major screens with dynamic 'Continue Journey' navigation.
• 100% Private Local Resume Analysis: Powered by Gemma 4 (gemma4:12b) on Ollama with strict local privacy boundaries. Raw resume data never leaves the student's device.
• Explainable Skill Gap Analysis: 3-category evaluation distinguishing Demonstrated Strengths (verified with concrete resume evidence), Mentioned/Partial Skills (coursework or keyword only), and Missing Job Requirements with status explanations.
• 104-Resource Curated Learning Catalog: Multi-modal technical materials spanning 4 dedicated categories:
  • 📺 WATCH: 45 verified YouTube video courses and tech conference talks.
  • 📖 LEARN: Official documentation, articles, and courses from Google Cloud, CNCF, and Python PSF.
  • 💻 PRACTICE: Hands-on interactive sandbox labs (Killercoda, Google Cloud Skills Boost, PGExercises).
  • 🛠️ BUILD: Canonical open-source architecture repositories (Google Cloud Microservices Demo, FastAPI, etc.).
• Compact Dark Resource Cards: Small, official 24–32px provider icons/favicons, difficulty and duration chips, and direct canonical action links (Watch Video ↗, Start Course ↗, Start Lab ↗, View Repository ↗) without large thumbnails or banners.
• Guided Weekly Learning Roadmaps: Strict 2-week or 4-week roadmaps with milestone checklists, expected deliverables, and evidence notes fields.
• Independent Progress Decoupling: Separate calculations for Weekly Completion %, Overall Learning Activity Progress %, and Career Preparation Tracker.
• Proactive 'Continue Learning' Banner: Highlights the next incomplete weekly milestone and its associated learning material.
• Exportable Career Action Report: Download comprehensive summary as an ATS-ready Markdown action plan.
• Clean Activity & Diagnostics: Student-facing milestone timeline paired with collapsed developer telemetry.

🧪 Automated Quality Gates & Verification

• Code Standards: Ruff linter and formatter passing 100% across 77 files.
• Test Suite: 75/75 pytest unit and integration tests passing.
• Browser Acceptance Suite: 10/10 automated Playwright browser workflows passing with strict heading assertions.
• Screenshot Manifest: 10 verified desktop views (1440 × 1000) documented in docs/images/SCREENSHOT-MANIFEST.md.

⚠️ Current Limitations & Roadmap

• Session-Bound Learning Progress: In v0.9.0, weekly checklist and evidence notes persist within active Streamlit session state; permanent SQLite/file storage arrives in v0.12.0.
• Resume Studio & AI Chatbox: Clarified as planned upcoming functionality (v0.10.0–v0.11.0); not yet active for student execution in v0.9.0.
• For complete details, see docs/known-issues.md.

v0.8.0 — OpenTelemetry & Grafana Observability

Milestone 8: OpenTelemetry + Grafana Observability Release

Highlights

• End-to-End Tracing: Instrument Google ADK Coordinator, Job Scout, Private Resume Agent, local Gemma 4 / Ollama, and Skill Builder with OpenTelemetry.
• Observability Stack: Complete Docker Compose profile (--profile observability) bundling OTel Collector, Tempo, Prometheus, and Grafana.
• Pre-provisioned Dashboards: Auto-provisioned Grafana dashboard at http://localhost:3000 with Prometheus metrics and Tempo trace search.
• Strict Privacy Boundaries: Automatic attribute sanitization ensuring raw resume text, user prompts, personal identifiers, completions, and API keys are NEVER exported.
• Verified Local Model Inference: Real end-to-end tracing validated against native Ollama running gemma4:12b (Apple Silicon Metal GPU accelerated) with captured token metrics and durations.
• Quality: 57 passing automated unit and evaluation tests; 100% clean Ruff linting and formatting; 100% automated browser acceptance test passing.

v0.7.0 — Dockerized Agentic Career Lab

Milestone 7: Docker-First Application Distribution

🚀 Overview

Milestone 7 makes Agentic Career Lab reproducibly deployable with a single command:

docker compose up -d

The application starts at http://localhost:8501, allowing students and engineers to upload resumes, analyze skill gaps privately with local Gemma 4 via Ollama, match jobs, and build personalized career roadmaps.

🌟 Key Capabilities

• Docker Compose First: Zero-configuration setup supporting local Gemma 4 inference.
• Production-Conscious Dockerfile: Multi-stage python:3.11-slim container executing as non-root appuser (UID 1000) with HTTP healthchecks.
• Hardware-Aware Ollama Architecture:
  • macOS / Apple Silicon: Default routing through http://host.docker.internal:11434 preserving native Metal GPU acceleration.
  • Linux / Standalone Profile: --profile standalone spins up a containerized ollama/ollama:latest instance with persistent Docker volume storage (agentic_career_lab_ollama_models).
• Air-Gapped Privacy Boundary: Strict isolation of raw resume text from persistent state and cloud services.
• All 53 Automated Tests Passing: Complete regression test suite and Ruff lint checks verified.

📋 Setup & Verification

See docs/antigravity/07-docker-deployment.md and docs/antigravity/m7-docker-acceptance.md for detailed architecture and measured test evidence.

v0.6.0 — Hybrid Gemini + Gemma 4 Working Demo

🎓 Agentic Career Lab v0.6.0 — Hybrid Gemini + Gemma 4 Working Demo

A privacy-aware, hybrid multi-agent career preparation platform for college students, combining local Gemma 4 reasoning, Google ADK orchestration, and Google Cloud Gemini planning.

---

🌟 Key Highlights & Features

• 🔒 Strict Local Resume Boundary (Gemma 4 on Ollama):

  • Private resume ingestion and evidence extraction run 100% locally on user hardware with gemma4:12b via Ollama (http://localhost:11434).
  • Raw resume text is never transmitted to cloud models (cloud_allowed=False, fallback_allowed=False).
  • Deterministic ClaimValidator validates suggestions against extracted evidence (0 unsupported claims accepted).

• ☁ Grounded Cloud Reasoning (Google Gemini on Vertex AI / API):

  • High-level learning roadmap synthesis and customized portfolio project generation powered by gemini-3.6-flash / gemini-3.8-flash.
  • Only sanitized metadata (target_role and missing skill names) is transmitted to the cloud via CloudPlanningContext.
  • Enforced by assert_cloud_safe() privacy guards.

• 🛡️ Explicit User Privacy Consent:

  • Interactive opt-in checkbox in the Streamlit UI before any structured data is shared with Cloud AI.
  • If consent is declined, execution automatically and seamlessly falls back to a deterministic Python offline planner.

• ⚙️ Google ADK Multi-Agent Orchestration:

  • ADKCareerCoordinator registers specialist agents (job_scout, private_resume_agent, skill_builder) under a unified coordination layer.
  • Produces a comprehensive, actionable CareerActionPlan with demonstrated strengths, priority gaps, weekly milestones, curated resources, and recommended portfolio projects.

• 🎯 Transparent Opportunity Sourcing:

  • Clear UI labeling for LIVE opportunities (Adzuna API) versus DEMO DATA (deterministic fixtures).

---

🧪 Test & Validation Evidence

• Pytest Suite: 53 passed in 0.69s (100% pass rate).
• Code Quality: Ruff lint and format clean (66 files checked).
• Live Local Gemma 4 Smoke Test: scripts/smoke_resume_local.py verified with live Ollama gemma4:12b (passed; zero secret leaks in workflow state).
• Live Cloud Gemini Test: Verified live generation via Google GenAI SDK with gemini-3.6-flash and gemini-3.8-flash.
• E2E Integration Verification: Full Job Scout → Private Resume Agent → Skill Builder → Career Action Plan workflow passed cleanly.
• Streamlit Health Check: Verified server initialization and HTTP 200 OK on /_stcore/health.

---

🚀 Installation & Quick Start

# 1. Clone repository
git clone https://github.com/hastimal/agentic-career-lab.git
cd agentic-career-lab

# 2. Setup Python environment
python3 -m venv .venv
source .venv/bin/activate
pip install -e ".[dev]"

# 3. Pull local Gemma 4 model via Ollama
ollama serve
ollama pull gemma4:12b

# 4. Configure environment (optional cloud key)
cp .env.example .env
# Set GEMINI_API_KEY for live cloud reasoning, or leave blank for deterministic Python fallback

# 5. Launch the Streamlit workspace
streamlit run src/agentic_career_lab/app.py

---

⚠️ Known Limitations

• Local Model Hardware: Local inference with gemma4:12b requires ~8GB RAM/VRAM. First-time model loading may take 30–60s depending on host machine hardware.
• Mock Fallback: If Ollama is offline, resume analysis halts with a clear error rather than leaking data to the cloud.

---

🗺️ Next Milestones

• Milestone 7: Docker containerization and Google Cloud Run deployment (uvx agentic-career-lab CLI packaging).
• Milestone 8: Evaluation benchmarks and OpenTelemetry observability.