FZAstro AI v2.3.1

FZAstro AI v2.3.1

This release focuses on the LLM pipeline refactor, polished tool output, Web Companion parity, weather/markets/news improvements, and release hardening.

Highlights

• Refactored the LLM request pipeline into shared modules for profiles, prompt building, context budgeting, request construction, and stream parsing.
• Cleaned up calibration/profile prompts and removed legacy prompt expansion.
• Added polished assistant rendering for links, tables, code blocks, weather summaries, market summaries, and structured app/tool replies.
• Added direct weather lookup using Open-Meteo, including current conditions and daily outlook.
• Added Global Market Pulse with indices, commodities, crypto, CRM, DBX, oil, gold, and broader global indicators.
• Improved Daily News output with cleaner cards, article summaries, source metadata, and better layout.
• Upgraded Web Companion so it uses the live frontend and shared app logic instead of the older embedded/basic implementation.
• Ported Web Companion SEEING and TARGETS to the same modern planners used by the desktop app.
• Fixed Web Companion SEEING internal-server-error handling.
• Added TARGETS result caching so repeated identical target searches return instantly after the first catalogue run.
• Improved UI polish across tabs, close buttons, Codex-style dark theme, tool menus, message spacing, and output rendering.
• Improved N.I.N.A. session flow and TARGETS handoff behavior.
• Expanded automated coverage for weather, markets, Web Companion tools, LLM request building, stream parsing, prompt profiles, context budgeting, and target caching.

Validation

• Full test suite passed: 356 passed
• Release build validation passed.
• EXE smoke test passed.
• Release artifact built successfully.

Artifact

• FZAstroAI.exe
• SHA256: 8A99B3E114B76C0204862E1F5D2FAA988EE9D625F419B580E406EBD887F85C5A

Notes

Tesseract OCR remains optional. If Tesseract is not installed or configured, OCR features are unavailable, but the main application and release validation still pass.

<style type="text/css"> p, li { white-space: pre-wrap; } hr { height: 1px; border-width: 0; } li.unchecked::marker { content: "\2610"; } li.checked::marker { content: "\2612"; } </style>

FZAstro AI v2.3.1 Help

This guide shows what FZAstro AI can do and which source/mode is usually used. Imaging Production v2.3 is the tabbed workspace release: chat stays clean, astronomy tools open as main-window tabs, the root folder has one deploy button, and release deploy can create the Git commit/tag automatically. Click the FZ square in the header to open the GitHub repository.

1. Main idea

Source chip	Meaning
LLM	The selected model answered from prompt/model reasoning.
Docs	The answer used the local Document Knowledge Library.
App	The app handled the request directly.
Web	The answer used internet search or webpage extraction.
News	The answer came from the Daily News workflow.
Market	The answer came from a structured market quote.
Memory	The answer used saved persistent memory.
Files	The answer used attached files.
Vision	An image-capable model inspected an image.
Python	Local Python code executed and returned output.

Good rule:

Ask normally for model knowledge.

Mention documents / PDF title / page number for local library work.

Mention web / latest / current / URL for internet work.

Mention remember / memory for persistent memory.

Mention run / test / execute for Python.

Use DEV for project scanning, coding context, plans, and checks.

2. Normal chat

Explain dithering in astrophotography.

Help me debug this Python error.

Compare OSC and mono cameras for deep-sky imaging.

Create a simple imaging checklist for tonight.

Expected source chip: LLM.

3. Models and telemetry

Use the model selector near the chat controls to choose the active local model. Use refresh after adding/removing Ollama models. If Ollama is installed but stopped on the default local endpoint, FZAstro AI can try to start ollama serve automatically.

The status line can show model name, response time, output size, context left, and token speed. The lower telemetry row can show GPU/VRAM, CPU, RAM, and best-effort temperatures.

4. Web, news, and market

Search the web for latest Artemis news.

Read this page: https://www.nasa.gov/blogs/artemis/

Summarize this article: https://www.nasa.gov/blogs/artemis/

Take screenshot https://www.nasa.gov/blogs/artemis/

Give me the daily news brief.

Get CRM stock price.

Get crude oil price.

Get gold price.

Expected source chips: Web, News, or Market.

5. Documents and files

Use attachments for small files you want the model to inspect directly. Use the Document Knowledge Library for large PDFs/books or long-term document search.

Analyze this attached Python file.

Import this PDF into the knowledge library.

Search my documents for telescope collimation.

Show page 12 from the PDF.

Expected source chips: Files, Docs, Vision, and/or LLM.

6. Python and Code Lab

Use Code Lab or ask directly:

Run this Python code.

Test this function with three cases.

Explain this traceback.

Create a pytest for this module.

Expected source chip: Python when code is executed.

7. LLM Benchmark Dashboard

Open Skills → Model Lab → LLM Benchmark. The legacy LLM BENCH wording remains documented for validation continuity.

The dashboard includes telemetry, Dashboard/History/Compare views, persona/calibration controls, Composite scoring, Run All Presets, and Delete Selected.

8. Astro Tools Suite

Use the workspace Apps button, Skills → Astro, or the Astro toolbar. The production tools are:

SITE, IMAGING, LOOKUP, SUN NOW, SEEING, TARGETS, and SOLAR MAP

• SITE — observing location, elevation, timezone, SQM, and Bortle context.
• IMAGING — camera preset, focal length, FOV, image size, and rotation.
• LOOKUP — object lookup with catalog details, sky preview, and distance-ladder information.
• SUN NOW — NASA/SDO solar imagery with metadata and cached fallback.
• SEEING — Astro Night Planner with current/tonight context, astronomical-dark priority, Moon periods, cloud-aware scoring, and Bortle tint.
• TARGETS — target ranking, filters, CSV export, and optional local OpenNGC import.
• SOLAR MAP — native 2D solar-system map with orbit/label/grid controls.

Bortle tint rules: 8–9 white/urban, 6–7 yellow, 4–5 green, 2–3 blue, and 1 violet.

Distance ladder calculations can use parallax, Gaia proxy, NED-D, and Hubble-style estimates where available. Optional visibility: FZASTRO_USE_DISTANCE_LADDER=1.

9. FZAstro Imaging / N.I.N.A. planning

Use ASTRO → FZASTRO IMAGING CONTROL or type safe predefined commands:

/nina-plan next

/nina-plan next 60s gain 200

/nina-plan target M13 60s gain 200

/imaging-plan target NGC 7000 exposure 120s gain 100 frames 80

The planner combines SITE, IMAGING, SEEING, and TARGETS context, then writes plan folders under:

Documents\FZAstroAI\Imaging Plans\<plan_id>

The file to open in N.I.N.A. Advanced Sequencer is the generated .nina-sequence.json. FZAstro can launch/open it for review when possible. The workflow is intentionally safe: it does not slew, center, guide, autofocus, capture, start a sequence, or schedule hardware execution automatically.

10. Web Companion

Use the desktop Web Companion controls or run:

.\scripts\run_web_companion.ps1 -Lan -Port 7860 -Token "fzastro"

Open the LAN URL from an iPad/phone/other device on the same network. LAN mode uses FZASTRO_WEB_TOKEN; do not expose it directly to the public internet.

11. AI Developer Workbench

Click DEV in the quick actions bar.

Use it to:

• Scan the project
• Select relevant files
• Build a context package
• Generate a visible implementation plan
• Run compile checks
• Run pytest checks
• Analyze failures
• Prepare safe patch workflows

Stage 1 is review-first: it helps prepare and validate coding work, not silently rewrite the project.

12. Useful commands

.\DEPLOY.bat

.\DEPLOY.bat -GitPush

python -m pytest

python -m compileall -q fzastro_ai tests

powershell -ExecutionPolicy Bypass -File .\scripts\validate_release.ps1 -PythonExe ".\.venv\Scripts\python.exe" -SkipLaunch

powershell -ExecutionPolicy Bypass -File .\scripts\deploy.ps1 -RunValidation -GitRelease