New Features

• HPC sliding-window submission (isce_base.py) — each step now runs a lightweight sbatch manager job that keeps ≤max_concurrent_hpc child jobs active at all times, refilling immediately on completion. Replaces the old batch-sequential approach. Consecutive steps with equal command counts (e.g. run_13–run_16) are merged into a single group-manager. Steps are chained via --dependency=afterok.
• Per-command elapsed time in sbatch logs (isce_base.py) — sbatch scripts print START/DONE/FAIL with elapsed seconds. Group tasks also print total elapsed across all grouped steps.

Bug Fixes

• HyP3 file paths (hyp3_base.py) — hyp3_jobs.json now saves to workdir root (was workdir/hyp3/); downloaded ZIPs go to workdir/hyp3/; retry job files save to workdir root; legacy out_dir=workdir entries auto-migrate to workdir/hyp3/.
• watch command ignores --interval flag (cli/main.py) — _proc_local_watch read args.refresh_interval but argparse stores it as args.interval; interval was always 60 s regardless of user input. Fixed.
• insarhub processor refresh/download re-processes retry job files (cli/main.py) — _find_job_files globbed hyp3*.json, matching hyp3_retry_jobs_<ts>.json files from past runs alongside hyp3_jobs.json. On refresh/download, stale retry files were loaded as separate processors, causing duplicate downloads or incorrect status. Fixed: retry files excluded from glob.
• Orbit files downloaded twice (cli/main.py) — -d -O flags triggered two orbit downloads. Fixed: skips explicit download_orbit() call when downloader already handled it.
• retry() runs locally after HPC submission (isce_base.py, cli/main.py) — hpc_mode excluded from saved config, so retry defaulted to local. Fixed: retry() auto-detects HPC from job metadata (slurm_job_ids/hpc_manager/hpc_array) and writes it back to config.hpc_mode so _step_executor() routes correctly. _load_local_processor also restores max_concurrent_hpc and HPC config fields from insarhub_config.json.
• Manager job killed mid-run leaves step stuck PENDING (isce_base.py) — elif n_cmds > 0 and not job_ids: was dead code (always False inside if job_ids: block); SLURM-killed managers with incomplete commands never resolved to FAILED. Fixed: condition is now elif n_cmds > 0:.
• _parse_time_s mis-parses 2-part SLURM time strings (isce_base.py) — "30:00" was treated as 30 h 0 min (108,000 s) instead of 30 min 0 s (1,800 s), overestimating group-manager walltime 60×. Fixed: 2-part strings now parsed as MM:SS per SLURM spec.
• Group-manager step stuck PENDING when job gone from SLURM (isce_base.py) — if group_task_dir was absent from saved metadata (old jobs), n_cmds resolved to 0, preventing the SUCCEEDED/FAILED transition. Fixed: n_cmds now stored in job metadata at submission; refresh uses it directly with file-count as fallback.
• refresh() shows only one RUNNING command in manager mode (isce_base.py) — with sliding-window, multiple commands run concurrently but only one showed RUNNING. Fixed: per-command status now derived from .done/.fail files; all in-flight commands show RUNNING.
• Hyp3_SBAS MintPy output in workdir root (mintpy_base.py) — base class run() passed self.workdir to TimeSeriesAnalysis; Hyp3_SBAS inherits without override, so all MintPy outputs scattered to workdir root instead of workdir/mintpy/. Fixed: uses self.mintpy_dir; same correction for _geocode_diagnostic_files.
• ISCE cleanup misses merged/interferograms (isce_sbas.py) — cleanup targeted isce/interferograms/ (nonexistent); real stackSentinel output is isce/merged/interferograms/. Large intermediate files were never deleted.
• HyP3 auth failure submits to wrong user (hyp3_base.py) — when re-auth failed for a pool user, credits=0 was overwritten by self.client.check_credits() on the previous user's client; jobs were then submitted under the wrong account. Fixed: credits check guarded by auth result.
• self.batchs updated per-loop-iteration (hyp3_base.py) — on multi-user refresh, if any user failed, their batch was silently dropped from self.batchs. Fixed: assignment moved after loop.
• Missing filename key in HyP3 file metadata crashes download (hyp3_base.py) — direct dict subscript raised KeyError for auxiliary entries lacking filename. Fixed: file_meta.get('filename') with skip on empty.
• delete_job_folder blocks on ~-prefixed workdir (settings.py) — Path(workdir) without expanduser().resolve() made relative_to() always raise ValueError, returning 403 on every delete. Fixed.
• _run_folder_select_pairs uses folder.parent as workdir (folders.py) — downloader config received parent directory instead of job folder; sub-paths written one level up, potentially colliding with sibling jobs. Fixed: workdir=folder.
• Analyzer stop_event leaks on step error (routes/analyzer.py) — early return on step exception bypassed _stop_events.pop(job_id), leaking events indefinitely. Fixed: pop before return.
• Refresh overwrites .insarhub_cache.json filenames with empty list (routes/processor.py) — if no jobs had SUCCEEDED yet, filenames=[] overwrote a valid cache from a prior successful refresh. Fixed: preserves existing filenames when current refresh yields none.
• Retry job files appear as selectable job entries in GUI (routes/processor.py) — hyp3*.json glob matched hyp3_retry_jobs_<ts>.json; selecting one for refresh returned only the retry batch status. Fixed: retry files excluded.

GUI

• Hyp3_S1 max_workers in settings panel (defaultconfig.py) — parallel download threads now configurable via Job settings group (default 4, range 1–16).

Refactor

• Centralized path layout (config/paths.py) — Hyp3Paths, ISCEPaths, MintPyPaths dataclasses replace all hardcoded workdir / "subdir" strings across hyp3_base.py, isce_base.py, isce_s1.py, mintpy_base.py, hyp3_sbas.py.
• Path dataclass coverage extended — remaining hardcoded path literals replaced with dataclass properties: isce_sbas.py now uses ISCEPaths for isce_dir, slc_dir, dem_dir; cli/main.py uses Hyp3Paths/ISCEPaths in _has_zips and _find_job_file; utils/batch.py and utils/tool.py use Hyp3Paths.output_dir for ZIP discovery.
• Hyp3Processor renamed to CloudProcessor (core/base.py) — ABC renamed to reflect generic cloud-backend semantics rather than HyP3 specificity. Updated across core/__init__.py, __init__.py, processor/hyp3_base.py, commands/processor.py, cli/main.py, core/engine.py.

Docs

• Contributing guide — new tab in MkDocs navigation (EN + ZH). Split into Overview, Backend, and Frontend pages.
• Backend contributing guide — architecture overview, path conventions, per-section instructions for adding new processors/downloaders/analyzers. Each section includes a "Adding a New Base X" subsection (with code examples for CloudProcessor/LocalProcessor, BaseDownloader, BaseAnalyzer) and an "Extending an Existing Base X" subsection with switch tabs (Hyp3Base, ISCE_Base, ASF_Base_Downloader, Mintpy_SBAS_Base_Analyzer).
• Frontend contributing guide — conda Node.js install, uvicorn backend startup from InSARHub root, module reference tables grouped by area (Entry & Global, Map, Search & Scene Selection, Jobs & Results, Settings, Utilities), backend communication pattern, settings panel, Vite proxy, build output, code style.

Bug Fixes

• Download result unpacking (commands/processor.py) — processor.download() returns (Path, dict) tuple; DownloadCommand was assigning the whole tuple to output_dir. Fixed: now unpacks to output_dir, dl_stats. CommandResult.data now includes both output path and download stats.
• HyP3 workflow marker wrong location (hyp3_base.py) — write_workflow_marker was writing insarhub_config.json to workdir/hyp3/ instead of the job folder root, so HyP3 tags never appeared in the job drawer. Fixed: writes to config.workdir.
• Stale out_dir from saved job file (hyp3_base.py) — old hyp3_jobs.json pointing to a pre-migration path outside current workdir would silently redirect output. Fixed: out_dir rejected if not under current workdir.
• wslpath unchecked (settings.py) — if wslpath -w failed, PowerShell was called with -File "". Fixed: returncode + empty string guard added.
• ZIP detection for hyp3/ layout (cli/main.py) — analyzer now checks workdir/hyp3/*.zip first, with fallback to workdir/*.zip for legacy layouts.

Performance

• Auth status parallel checks (auth.py) — HyP3 credit check, CDSE, CDS, and Earthdata checks now run concurrently via ThreadPoolExecutor instead of sequentially. Typical improvement: 3–5× faster settings panel load.
• Job folder listing SSH speed (settings.py) — removed all per-folder glob/exists/is_file checks. Now reads only insarhub_config.json per folder. Significant speedup on remote filesystems.

Source

• hyp3/ subdir awareness (hyp3_sbas.py, mintpy_base.py, batch.py) — all ZIP lookups now check workdir/hyp3/*.zip first, falling back to workdir/*.zip for legacy layouts. Affected paths: _unzip_hyp3, cleanup, and ERA5Downloader.download_batch.
• Missing .mintpy.cfg guard (mintpy_base.py) — if .mintpy.cfg is not found when run() is called, a warning is printed and the config is written automatically rather than crashing downstream MintPy steps.
• write_mintpy_config parent mkdir (defaultconfig.py) — outpath.parent.mkdir(parents=True, exist_ok=True) added before opening the file, preventing FileNotFoundError when the output directory does not yet exist.

CLI

• prep alias — insarhub analyzer run --step prep now accepted as alias for prep_data. Help text updated to show alias.
• Default port — insarhub-app now defaults to 8080 (was 8000). Use --port to override.

GUI

• Subfolder navigation (JobQueueDrawer) — click any folder to drill into subfolders; ↑ button to go up. Resets to workdir root on workdir change. Uses /api/browse-subfolders endpoint.
• Cancel button (JobQueueDrawer) — Cancel action added for local ISCE jobs.
• Modern folder picker (settings.py) — Windows/WSL now uses IFileOpenDialog COM API via embedded C# in PowerShell. Fixes: DPI blurriness on 2K monitors, Chinese character paths.
• Add nyan cat

Network Graph (utils/tool.py)

• Node labels changed from last-8-chars to YYYY-MM-DD dates.
• Bottom axis: real acquisition dates. Top axis: days since first acquisition (swapped).
• Left graph title removed.
• Font sizes increased throughout; date labels rotated for readability.

Docs

• Port references updated to 8080 across README, quickstart, and frontend docs.
• file_structure.md/zh: added hyp3/ to directory layout; out_dir examples updated to .../hyp3.
• cli.md/zh: --credential-pool corrected from "JSON" to plain username:password text file; prep → prep_data.
• index.md/zh: satellite support table added; program structure section moved to new Advanced page with workflow diagram.

V0.3.0 ISCE2 Local Processor Support/HPC support

Changelog

[0.3.0] - 2026-05-14

New Features

• ISCE_S1 local processor: New processor backend that runs ISCE2 stackSentinel locally. Supports sequential local execution and SLURM HPC mode (--hpc-mode). Bounding box is auto-filled from the map AOI in the GUI.
• ISCE_SBAS analyzer: New MintPy SBAS analyzer for ISCE2 stackSentinel outputs. prep_data() auto-discovers interferogram, geometry, baseline, and metadata paths; MintPy outputs written to mintpy/ subdirectory.
• HPC mode (SLURM): ISCE_S1 can submit each processing step as a separate sbatch job. Per-step resource configuration via sbatch_options.json, editable in the GUI via Sbatch Options modal.
• Job Folders subfolder browser: The Jobs drawer now lists both folders and files. Click any folder to navigate into it; click ↑ Up to return to the parent. Breadcrumb path shown in the header.
• Cancel button for local processors: A Cancel button appears in the ISCE_S1 processor panel to terminate the running background process (local) or scancel all active SLURM jobs (HPC).
• Refresh with per-command detail: ISCE_S1 refresh() now shows per-command status (cmd_NNNN RUNNING / SUCCEEDED / FAILED) for multi-command steps, matching the CLI output.

Bug Fixes

• ISCE_S1 bbox not passed: Processor.create() was calling cls(cfg) which mapped the config to the pairs argument in ISCE_S1's two-argument constructor. Fixed by detecting pairs in the constructor signature via inspect.signature and using keyword arguments.
• ISCE_SBAS diagnostic geocoding: avgPhaseVelocity.h5, numTriNonzeroIntAmbiguity.h5, and maskConnComp.h5 are now geocoded automatically after the geocode step. Existing radar-coordinate data is geocoded on demand in the render endpoint.
• ISCE_SBAS timeseries filter: View Results now returns only geo/geo_timeseries*.h5 (geocoded) when present, not the radar-coordinate timeseries*.h5 files.
• ISCE_SBAS .mintpy.cfg path: Analyzer route was writing .mintpy.cfg to the job folder root; ISCE_SBAS expects it at mintpy/.mintpy.cfg. Fixed by reading analyzer.cfg_path at runtime.
• ISCE_S1 submit via GUI missing sbatch options: _run_folder_process now loads sbatch_options.json and calls processor.submit() directly for local processors, bypassing the HyP3-only SubmitCommand/SaveJobsCommand wrappers.
• cmd index parsing crash: int() raised ValueError on malformed cmd_????.done/fail filenames. Fixed with a safe _idx() helper.
• Job Folders empty workdir path traversal: An empty workdir in browse-subfolders resolved to CWD, allowing requests outside the workdir. Fixed with an early 400 response when workdir is not configured.
• Job Folders has_children OSError: subfolder.iterdir() on restricted directories could raise OSError. Wrapped in try/except.

Full Changelog: v0.2.5...v0.3.0

Key updates:

1. Network scoring system and modification before interferogram generation. Based on S1 global coherence dataset and weather data
2. add quick_overview in mintpy for quality control
3. re-oragnize fastaoi into seperate sub modules

Pair Quality Database:

• DB no longer rebuilds when pair selection parameters change (dt_max, pb_max, min_degree, etc.) — only rebuilds when the scene set actually changes
• Removing scenes from the network no longer triggers a rebuild; the existing DB already covers all subsets
• DB now stores _scene_names for exact scene-set comparison instead of relying on count alone
• Backward compatible: old DBs without _scene_names fall back to count comparison and migrate on the next rebuild
  Performance
• S1 global coherence prefetch is now parallelized in two phases: season tile S3 downloads run concurrently (up to 4 threads), followed by per-pair numpy coherence evaluation in parallel (8 threads). Expected speedup for 32,000+ pairs: 4–6× on first run, with warm cache unchanged

select_pairs: Avoid Bad Acquisition Days (avoid_low_quality_days)

• Default changed to True — bad-weather scenes are excluded from the network by default
• Default precipitation threshold tightened to 25 mm (3-day accumulation)
• Weather and snow data fetched during bad-scene filtering is now seeded directly into the pair quality cache — the scorer reuses it instead of making a second round of API calls, eliminating duplicate fetches entirely
• select_pairs returns a prefetch_cache containing the pre-fetched weather/snow keyed by stack, which _run_folder_select_pairs writes into each subfolder's CacheManager before scoring begins
• Status message "Building weather/snow cache — P{path}/F{frame}…" shown in the GUI while seeding so the user can see progress during long cache writes
  main
  ·
  v0.2.5
  1 paren

Full Changelog: v0.2.4...v0.2.5

[0.2.4] - 2026-03-25

New Features

• CLI & API: select_pairs() is now a pure computation method — no file I/O inside the class. File writing (JSON, PNG, workflow marker) has been moved to the CLI and API call sites, keeping the core logic reusable and testable
• Path handling: All functions that accept path arguments now call .expanduser().resolve(), enabling ~ tilde paths everywhere
• WebUI: Added documentation button in the General Settings panel (bottom-left) linking to the InSARHub docs site
• WebUI (insarhub-app): Auto-creates the working directory if it does not exist when -w <path> is passed
• CLI (insarhub-app): Added -v / --version flag
• Windows fix: insarhub-app no longer returns immediately on Windows — sets WindowsSelectorEventLoopPolicy so uvicorn blocks correctly

Bug Fixes

• WebUI Processor: Unchecking dry-run after a completed run no longer leaves the button stuck at "✓ Done" — the status resets to idle on checkbox change
• WebUI Processor: Clicking "✓ Done" after a real (non-dry-run) submit now correctly closes the modal
• WebUI Processor: "✓ Done" button now shows a pointer cursor on hover
• Analyzer: Fixed NoneType crash in troposphere correction when Path.mkdir() was called on an already-resolved path
• CLI credential setup: Removed spurious blank first line from .cdsapirc written by the interactive credential prompt

v0.2.3

Bug Fixes

• Fixed troposphere correction crash — Path.mkdir() returns None, which was incorrectly assigned and passed to pyaps3 as the weather data directory
• Fixed .cdsapirc file created by CLI prompt having a blank first line
• Fixed dry run checkbox leaving the submit button stuck at "✓ Done" after job completion
• Fixed "✓ Done" button not closing the modal after a real (non-dry-run) submission
• Fixed missing pointer cursor on "✓ Done" button
• Fixed pixel hover value lookup to use Mercator projection formula instead of linear WGS84 interpolation
• Fixed unclosed file handle in /api/folder-image endpoint
• Fixed RuntimeWarning: invalid value encountered in cast when rendering velocity PNG with NaN pixels
• Added path traversal protection to /api/folder-image endpoint
• Fixed polling interval leak in frontend job status watcher

Documentation

• Home page: corrected package name from "InSAR Script" to "InSARHub", added Web UI overview screenshot and button, clarified WSL2 support
• Header: added Email and Discord icon buttons next to GitHub badge
• Header: reduced excess whitespace in GitHub badge

Features

• Added yellow click marker on the map when a time-series pixel is selected
• Click marker renders above the velocity overlay layer

v0.2.1

Web UI

• Velocity map alignment fixed — corrected a projection error where UTM-projected MintPy outputs were reprojected to EPSG:3857 using wrong source CRS, causing the velocity overlay to appear shifted relative to the basemap
• Time-series click marker — clicking a pixel on the velocity map now shows a yellow highlight point to indicate the selected location
• Per-job analyzer status — the Stop button and run status in the Analyzer panel are now isolated per job folder; switching folders no longer bleeds status across jobs
• Drawer auto-hides on map click — the job drawer collapses automatically when clicking on the map to extract a time-series

Downloader

• Added --orbit-files option to download precise orbit files alongside scenes
• Added granule name input: users can now supply a file, list, or string of granule names for custom scene searches
• Fixed --stacks with multiple entries (e.g. 28:107 28:112) being incorrectly deduplicated by the ASF server
• Orbit download now falls back to ASF server automatically if the CDSE server is unavailable
• Fixed duplicate EOF file detection by checking acquire time

Bug Fixes

• Fixed download error not enough values to unpack in the parallel download result handler
• Fixed NumPy warning invalid value encountered in cast when rendering velocity colormaps with NaN pixels
• Pinned CI to Python 3.12 to avoid instability with 3.13/3.14

Documentation

• Completed Web UI documentation: Search & Download, Processor, Analyzer, and Results Viewer pages
• Added figures for all UI panels with light/dark mode variants

Changelog

0.1.0 - 2026-03-06

Initial Release

First public release of InSARHub — a modular Python framework for automated InSAR time-series processing.

Features

Downloader

• ASF_Base_Downloader: Search and download Sentinel-1, ALOS, and NISAR SLC data via the ASF Search API
• Spatial filtering with bounding box, WKT, or GeoJSON/shapefile AOI
• Post-search filtering by date range, path/frame, flight direction, polarization, season, coverage, and scene count
• Scene footprint visualization with basemap overlay (footprint())
• DEM download via dem-stitcher aligned to search footprints
• Multi-threaded download with Ctrl+C cancellation and partial-file cleanup
• S1_SLC: Sentinel-1 SLC specialized downloader with orbit file (sentineleof) support

Processor

• Hyp3_InSAR: Submit, monitor, download, retry, and persist HyP3 InSAR jobs
• Multi-account credential pool with automatic credit-aware job rotation
• Batch job persistence (save/load JSON) for resumable workflows
• watch() mode: polls job status and downloads succeeded outputs continuously
• Retry failed jobs with automatic timestamp-stamped save files

Analyzer

• Hyp3_SBAS: End-to-end MintPy SBAS time-series analysis from HyP3 outputs
• Automatic unzip, file collection, common-overlap clipping, and MintPy config generation
• Optional pyAPS tropospheric correction with CDS API credential management
• cleanup() to remove temporary files after processing

Utilities

• select_pairs: Temporal and perpendicular baseline filtering with configurable targets and tolerances
• Local baseline computation (zero network calls for Sentinel-1 and ALOS)
• API fallback with threaded fetching for products without local baseline data
• Connectivity enforcement: minimum/maximum degree per scene with force-connect option
• plot_pair_network: Network visualization with per-scene connection histogram
• ERA5Downloader: Batch ERA5 reanalysis download for MintPy tropospheric correction, MintPy-compatible filenames
• clip_hyp3_insar: Clip HyP3 zip outputs to a custom AOI before analysis
• Slurmjob_Config: Generate SLURM batch scripts for HPC job submission
• earth_credit_pool: Load multi-account Earthdata credentials from a pool file

CLI (insarhub)

• insarhub download — search, filter, and download SLC scenes
• insarhub processor submit/refresh/download/retry/watch/save/credits — full HyP3 job lifecycle
• insarhub analyzer prep/run — prepare and run MintPy analysis
• insarhub utils select-pairs/plot-network/era5/clip — utility commands
• Workdir (-w) and credential pool (--credential-pool) flags across all subcommands

Core

• Auto-registering component registry (Downloader, Processor, Analyzer)
• InSAREngine: high-level pipeline runner with skip flags and watch mode
• Unified CommandResult pattern shared between CLI and Panel frontend