All documents

This page is generated from the repository on every merge to main; changes made here are overwritten.

The repository keeps its history in documents. This index lists them all and links each to GitHub. Pages that are living references are also mirrored on this site (see the sidebar); everything else stays as the original file so it remains an accurate record of when it was written.

How to read the sections:

• Living references are updated as things change.
• Reports, plans, sheets and audits are accurate to the date they were written. They are never edited afterwards; a newer document supersedes them.
• The archive holds early specifications and reviews, frozen.

The 36-section task register and the tool reference are generated from the code; see How this site is built.

Living references

Kept current as the code changes.

Document	What it is
Cartogen AI — Bug Tracker	Lightweight, in-repo, git-versioned bug tracker. Created 2026-08-21 as part of setting up ongoing implementation tracking — this is the first version,…
Full live test: rc34 (every category, every input, every output)	This is the one document to follow for a complete hand test of the installed plugin. It exists because the last full hand run was on rc22 (2026-10-09,…
Humanitarian Cartography Standards	This document records the compliance standard the project owner supplied on 2026-09-03 ("Comprehensive Humanitarian Operations and Mapping Standard Fr…
Humanitarian Mapping Task Reference	This reference taxonomy defines the humanitarian mapping work Cartogen AI should support. It is used to guide tool coverage, prompt examples, workflow…
Humanitarian tools catalogue	Every tool Cartogen AI offers for humanitarian work, organised by the six workflows in the "Humanitarian Mapping Workflows" document (field operations…
Cartogen AI — Implementation Tracker	This is the one place to look for "what's actually still open right now." Every review, audit, and spec doc referenced below now lives in docs/archive…
Cartogen AI — Live Test Scenarios	Log results the same way RELEASESMOKETEST.md does: append a dated entry to that file's Run log section noting which of these 5 scenarios passed/failed…
Live smoke test: rc37 (start here)	This is the short guide. It adds 23 rows (RC37-) for what changed after rc34 (rc35 to rc37) to the existing full pack, and tells you which of the olde…
docs/ index	Living references and trackers for the Cartogen AI QGIS plugin, grouped the same way as the main README's Documentation table — the two are kept in sy…
Cartogen AI — Interactive Release Live-Test Scenarios	This runbook expands Part 2 of RELEASESMOKETEST.md into an end-user test covering all 16 functional categories. Run it from the installed candidate ZI…
Cartogen AI — Release Smoke Test Checklist	Use RELEASELIVETESTSCENARIOS.md for the complete operator procedure. It expands the 16 categories below into three end-user scenarios with candidate-Z…
Tool Reference	Auto-generated from the live tool registry (335 tools) by docs/generatetoolsreference.py -- do not hand-edit, regenerate instead so this can never dri…
User Guide	Open Cartogen AI from the toolbar icon or the Plugins menu. It's a dock panel (drag it to float, or dock it left/right) — one continuous scroll, not s…

Reports, plans, sheets and audits

Each is accurate to the date it was written (most carry a date or a release candidate number in the name) and is never edited afterwards; a newer document supersedes it.

Document	What it is
Acceptance run sheet: cost-and-routing branch (PR #227)	Purpose: the first real test, with a real model, of what PR #227 added: the cost measurement, the loop guard and token budget, the job queue, runsteps…
API call optimisation plan -- 2026-10-10	Status: plan, nothing in section 4 is built yet. Section 2 is measured on main (63ceace+) with the repo's own prompt builder and tool router; section…
Deep code audit, 2026-10-28	Dated snapshot (do not edit later; supersede with a new doc). Measured on main at 75e719c, before PLAN-040 stage 0/3/4 merged (#485 c34ba21, #486 aaf6…
Remote branch prune: dry run, 2026-11-09 (HY-4)	Status: dry run only. Nothing was deleted. Deleting a remote branch cannot be undone from here, so the owner approves this list first.
Broad exception handlers: review snapshot, 2026-11-09 (DEF-8)	A dated snapshot (not edited later). Produced by python tools/broadexceptreview.py (new, standard library only); tests/testbroadexceptreview.py pins t…
Cartogen AI (cartogen-ai) — Full Code Review, 2026-09-08	Independent 5-dimension review (mockups/incomplete UI, bugs, uncompleted work, security vulnerabilities, GDPR compliance) of this repo (C:\Cartogen-AI…
Cost, routing and missing-capability plan (revised 2026-10-09)	Trigger: a user pasted five humanitarian scenarios (Marib health access, Hadramawt/Al Mahrah district areas, Aden water service gaps, Taizz terrain, c…
Project status dashboard: how it is built and refreshed	A private Claude Artifact (a "Dashboard" page) shows the project's status: version and counts, how much has been checked by hand, the latest hand-test…
Analysis of the nine "Decision needed" rows, 2026-11-09	Written at the owner's request after the master list was re-read. It is a decision aid: nothing here has been applied except the status fixes listed a…
execute_pyqgis_script process isolation — architecture scope	Every fix to date (os/sys/cartogenai blocks, QgsProject.write()/authManager() blocks, the type.dict and frame-walking closures, the 2026-09-04/09-05 s…
Prompt for an AI agent that runs the full live smoke test	Paste everything below the line into the agent. It assumes an agent that can drive a desktop (QGIS on screen, a terminal, the file system). If the age…
Failure protocol - what to capture the moment a row fails or looks wrong	Read this once before row 1. It adds to EVIDENCEHOWTO.txt (which says how the routine evidence is collected); it does not replace it. "Pass" or "Fail"…
GDPR technical addendum -- Hosted-Account feature (agent/account.py / ui/account_dialog.py)	This addendum exists to close the informational half of BUG-2026-09-08-2: correction, 2026-09-08 -- this paragraph originally said the review predated…
Hand-test package for the 1.18.5 features (rows W1 to W12)	For whoever runs the test: a person at QGIS, or an AI agent driving QGIS and a terminal for them. The rows themselves (exact text to type, files in, f…
Results of the 1.18.5 hand-test run, 2026-11-03 (triage)	Run by an AI agent driving QGIS 4.2 and a terminal on a Windows machine (provider: a Codex CLI client), not by a person. Source: results/4RESULTS.csv…
Full live smoke run, 2026-11-04: triage of the agent's results	Run by an AI agent driving QGIS 4.2.3 on Windows 11 with Gemini (gemini-flash-latest), package cartogenaifullsmoke1.18.5 (44 rows). Reported: 5 Pass,…
Hand test results, 1.17.0-rc48 (run 2026-10-23 and 2026-10-24)	Source: 4RESULTS.csv of the rc48 bundle, filled in by the owner. Plugin zip SHA-256 9d7a0ce1232f4e84063fcef12d440367c1bd800b0fa8d497bef22129de39fc88 (…
Plain-language algorithm test (humanitarian GIS), 2026-10-17
Hand-test sheet for everything new in 1.17.0 (rc42 to rc44; run on the rc46 build) -- 2026-10-16	def mem(kind, name, fields, rows): lyr = QgsVectorLayer(f"{kind}?crs=EPSG:4326&" + "&".join(f"field={f}" for f in fields), name, "memory") feats = []…
Humanitarian workflow gap analysis (2026-10-04)	Source: the owner's document "Humanitarian Mapping Workflows: From Field Operations to Strategic Orchestration" (6 workflows in two phases). Compared…
Hydrology engineering tools (watershed / peak flow)	The tools were written in the owner's local copy of 1.16.0-rc12 after a live request,
Watershed delineation (showcase S1), 2026-10-14	This supersedes the "does not delineate a basin" limit in docs/HYDROLOGYENGINEERINGTOOLS2026-10-04.md. That document is a dated record and is not edit…
Structured LLM output (Instructor/Outlines) — not recommended; a cheaper native fix exists	core/services/promptrefiner.py has two call sites that ask the model to produce JSON inside a plain-text system prompt instruction ("Respond as JSON o…
Integration audit: do the new tools work with the old ones? (2026-10-16)	Question: after packages A to I and the gap-closing work (42 new tools, about 160 new allow-listed Processing algorithms), is routing correct, are the…
ISSUE TRIAGE AND PLAN 2026-10-04
Issue verification checklist (open issues, 2026-10-05)	A dated checklist, not a result. Nothing here has been run by hand. It turns the open GitHub issues into hand-test rows so each can be closed on a pas…
JIAF 2 analysis-support module -- stage 1 plan and specification (2026-10-04)	Sources, all supplied by the project owner:
Vector memory (LanceDB/Chroma) for memory.py — scoped, not recommended yet	SpatialMemoryManager stores three kinds of state: per-project notes (storeprojectnote, opt-in persistence — GDPR finding F6), global notes (storegloba…
Lean hand-test protocol (from rc47)	Why: the rc46 run re-tested ~245 algorithm rows whose code had not changed since rc45 and cost a lot of API credit. Every live row sends about 24k tok…
Map profiles	A profile is a small JSON file that holds the choices your team repeats on every map. Name it in a request ("make a situation map with the profile wee…
Master task list (2026-10-27, after release 1.18.4)	One list to follow up and plan from. It replaces the session task list (101 entries: 84 done, 17 open) and consolidates the open work of PLAN-022, PLA…
Cloud-provider egress gate (the "Ollama-only" policy) — scope	1. Pre-dispatch in realexecutetool (the same place as the plan gate): resolve layer-name arguments → tags + lineage → block before the tool runs. 2. R…
Open issues: what the code does today (2026-10-06, after rc19)	Supersedes the "Still open inside it" column of docs/ISSUEVERIFICATIONCHECKLIST2026-10-05.md (that file is a dated snapshot and is not edited). "In co…
Open GitHub issues: status at rc35 (2026-10-11)	A dated snapshot, not a result. 46 issues are open. None was closed in this pass, because the rule is that an issue closes only on a hand-verified pas…
Open issues, status on 2026-10-23 (rc47)	Source: the last comment on each issue (most are from the rc40 hand test, 2026-10-14) and the rc41 to rc47 changelog. rc41 to rc47 were shipped after…
Open issues, status on 2026-10-24 (rc48 hand test)	Source: the rc48 hand test (docs/HANDTESTRESULTSrc482026-10-24.md, 87 rows run by the owner on 2026-10-23 and 24) and the earlier comments. Policy: no…
Open issues, status on 2026-10-25 (after the rc50 partial hand run and the rc51 release)	Supersedes OPENISSUESSTATUS2026-10-24.md (kept as written). Policy: no issue is closed without a hand-verified PASS or an owner decision. The hand-tes…
Open issues, status on 2026-10-25 (after PLAN-021, with 1.18.1 built)	Supersedes OPENISSUESSTATUS2026-10-25.md (kept as written; the hand-test bundle builder of 1.18.0 reads it). Policy until today: no issue closed witho…
What is left, and only you can do it (2026-10-12, build rc37)	A dated checklist, not a result. Everything that can be done from the repository was done; each item below needs your account, your data or a desktop…
Full review and owner queue, 2026-11-07	A dated snapshot (not edited later). It reads the master task list (101 rows), the live-results record, the open issues and the round-2 evidence as of…
All pending work, and the core analysis by professional profile (2026-10-26)	A dated review, not a result. It was written at the owner's request ("review all documentation and make a comprehensive list of all pending work, and…
Persona catalogue	Generated by python docs/generatepersonacatalogue.py from src/cartogenai/core/personas/.json and the task register. Do not edit by hand: tests/testper…
Persona framework: tools, data, styling and visualization plan (2026-10-04)	Source: the owner's "QGIS System Personas" document (15 personas: tasks, GIS functionalities, online data sources). Compared with what is in this repo…
Outbound PII scanning (Presidio) — scoped, not recommended	docs/OLLAMAENFORCEMENTGATESCOPE2026-09-24.md §4 lists four candidate mechanisms for the egress gate and explicitly names this one:
Processing-cost audit: what freezes QGIS, and what was done (2026-10-12)	Owner report: several freezes and hangs; wanted better queue and memory management, and big processing split into smaller functions.
QGIS coverage gap analysis (2026-10-14)	Owner request: read the QGIS and plugin-developer documentation, list every function, vocabulary item, action, Processing tool, QGIS feature, layer ty…
QGIS Production Architecture Review — 2026-09-04	This document records a 27-point architectural review Baron supplied on 2026-09-04 proposing a "QGIS-first production standard" for Cartogen AI -- cen…
rc10 – rc12 smoke-test audit (2026-10-02)	Self-review of every finding from the hands-on smoke tests of rc10 and rc11 (Yemen / Sanaa, QGIS 4.2.2), and what rc12 does about each. Dated snapshot…
Cartogen AI v1.16.0-rc10 — Interactive Smoke Test (Yemen / Sanaa reference)	Operator guide for the rc10 candidate. It replaces RELEASESMOKECHECKLISTrc7YEMEN.md for this round and covers:
rc12 live test and audit plan
Installed RC15 live smoke rerun — 6 October 2026	The real QGIS desktop and installed Cartogen AI chat were used. No repository tool imports or headless QGIS calls substituted for the live tests. Succ…
rc15 installed-profile smoke report: triage (2026-10-06)	A dated record, not a result of re-testing. Input: docs/RC15LIVESMOKEREPORT2026-10-06.md (the report as received; screenshots and artifacts stayed wit…
Cartogen AI RC17 live smoke test — 2026-10-06	Status: COMPLETE — OVERALL FAIL for the installed RC17 functional smoke run. Seven categories passed their core checks, eight failed a required subche…
rc17 installed-profile smoke report: triage (2026-10-06)	A dated record, not a re-test. Input: docs/RC17LIVESMOKEREPORT2026-10-06.md (as received). Of the 7 regression rows, 5 passed their core check (R1, R2…
Cartogen AI RC18 live smoke test — 2026-10-06	Status: IN PROGRESS. User installed RC18 and authorized proceeding; Plugin Manager verified 1.16.0-RC18 (01verifiedrc18.png). QGIS was closed and rela…
rc18 hand test: triage (2026-10-06)	Source: docs/RC18LIVESMOKEREPORT2026-10-06.md (as received, status "in progress"), with its screenshots and logs. Nothing here is hand-verified by me;…
RC22 live smoke test — current results, 2026-10-09	Status: IN PROGRESS — Gemini initially remained blocked after the credit top-up, then returned CREDITOK at 12:28 local time. Provider-dependent execut…
rc22 live smoke test: triage and fixes (2026-10-09)	Source: the owner's hand run of the rc22 zip (SHA-256 44bd1742...8414c) on QGIS 4.2.3 with Gemini (gemini-flash-latest), saved as docs/RC22LIVESMOKERE…
rc24 hand-verification sheet	Zip: cartogenaiv1.16.0-rc24.zip from the cartogen-ai-v1.16.0-rc24 pre-release (check it against SHA256.txt on the release). rc24 contains everything i…
rc34 hand-verification sheet	Zip: cartogenaiv1.16.0-rc34.zip from the cartogen-ai-v1.16.0-rc34 pre-release. Check it against SHA256.txt on the release page before installing. Use…
rc37 live smoke test: results and triage (2026-10-13)	Run on the rc37 bundle in desktop QGIS 4.2 with Gemini (gemini-flash-latest), by a computer-use agent following 3PROMPTS.html, with the plugin's own e…
rc44: what was implemented, and what the QGIS 4.2.3 reference check found (2026-10-16)	Source of the check: the "QGIS 4.2.3 Developer Reference" companion (PyQGIS 4.2 API pages with signatures, 465 Processing tools, 466 expression functi…
rc46: what the rc45 smoke run found, what was changed, and how far each change is checked (2026-10-17)	Source: the filled results of the rc45 hand-test sheet (an AI agent drove QGIS 4.2.3 on Windows with the plugin using Gemini; see docs/RC44STATUSANDAP…
v1.16.0-rc7 — Interactive smoke test: results, findings, decisions	Severity: high = wrong result / data loss / security-flow / dishonest output; medium = misleading or costly; low = cosmetic. "Status" is as of this sn…
rc8 audit, researched solutions and long-term recommendations (2026-09-30)	Dated snapshot; supersedes the status wording in RC8ISSUEREVIEW2026-09-30.md where they differ. Rules I held myself to: every claim says how it was es…
rc8 fix plan: all 25 findings + umbrella #72 (2026-09-30)	Dated plan. Builds on RC8AUDITANDSOLUTIONS2026-09-30.md. Constraint from the owner: no new external service or API. Every action below uses only QGIS/…
rc8 issue review: F01-F25 + umbrella #72 (2026-09-30)	Dated snapshot (not edited later). Basis: main at c0de3f5, the rc8 zip built from it, CI green on that head (offline 2,537 tests; live-QGIS job 97 tes…
Cartogen AI v1.16.0-rc7 — Interactive Smoke Checklist (Yemen / Sanaa reference)	Operator checklist for the rc7 candidate. It layers two things:
execute_pyqgis_script — RestrictedPython as a defense-in-depth layer	pip install RestrictedPython (v8.5) and its real compilerestricted/safeglobals/ safergetattr were run directly against this exact codebase's own histo…
Roadmap audit: the ten proposed features against the code (2026-10-27, build 1.18.4)	Scope: a strategic list of ten features and a competitor matrix was proposed. This audit reads the repository at main (1.18.4) and states, for each fe…
Completion plan for the open roadmap items (2026-10-16)	Supersedes nothing: docs/ROADMAPQGISCOVERAGE2026-10-14.md stays the record of what was agreed. This note orders the work that is still open after M0,…
Roadmap: full QGIS control through the chat (2026-10-14)	Plain-language plan. Behind it are docs/QGISCOVERAGEGAPANALYSIS2026-10-14.md (the evidence) and the two spreadsheets in docs/qgiscoverage/.
Router regression audit, 2026-10-14	The first pass checked only short requests and only which tools are offered. Each gap now has a deterministic check, committed, and run in CI.
Runbook expectation checks (DEF-16), 2026-10-28	DEF-16 was "runbook expectations to correct": rows where the 1.18.4 hand run failed because the runbook said the wrong thing, not the tool. The runboo…
Self-audit and harmonization, 2026-10-23	Written at the owner's request: check the project's instructions and skills, audit my own work against them, and make the pieces work together so that…
tool_router.py's top-40 filter — semantic-router prototype, not a build plan	filterrelevanttools(userquery, topk=40) is a candidate-set filter, not final tool selection: word-boundary keyword matching against each tool's name a…
Showcase services: what was measured, and from where (2026-10-16)	The S1 to S4 showcase fixes depend on four outside services. Until now they were labelled "unverified". tests/servicereachcheck.py calls each one with…
Live smoke-test run sheet — rc15 (prepared 2026-10-05)	A blank sheet, not a result. Nothing here has been run. It orders the hand checks from docs/ISSUEVERIFICATIONCHECKLIST2026-10-05.md (issue rows), docs…
rc16 re-test sheet (prepared 2026-10-06)	A blank sheet, not a result. Nothing here has been run. It covers only what the rc15 hand test found; after it passes, continue with docs/SMOKERUNSHEE…
rc17 re-test sheet (prepared 2026-10-06)	A blank sheet, not a result. Nothing here has been run. It re-tests what the rc15 report found and what rc16 and rc17 changed. Use the smoke fixture (…
rc18 live smoke test sheet (prepared 2026-10-06, expanded the same day)	A blank sheet, not a result. Nothing here has been run. It covers (1) the seven rc17 regression rows, (2) every fix made after the rc15 and rc17 repor…
rc19 live smoke test sheet (prepared 2026-10-06)	A blank sheet, not a result. It re-tests the fixes made after the rc18 hand test (docs/RC18SMOKETRIAGE2026-10-06.md) and carries over the rows the rc1…
rc20 live smoke test sheet (prepared 2026-10-06)	A blank sheet, not a result. rc20 adds the audit fixes and CI coverage made after rc19 (see the v1.16.0-rc20 block in metadata.txt). The rc19 sheet (d…
rc21 live smoke test sheet (prepared 2026-10-07)	A blank sheet, not a result. rc21 fixes three things the rc20 hand test found (see the v1.16.0-rc21 block in metadata.txt). The rc20 sheet (docs/SMOKE…
rc22 live smoke test sheet (prepared 2026-10-07)	A blank sheet, not a result. rc22 carries the fixes from the rc20 architectural audit (see the v1.16.0-rc22 block in metadata.txt). The rc21 sheet (do…
Frozen snapshots in docs/	Generated by python tools/docsindex.py; do not edit by hand. Each document below is accurate to the date or release in its name and is never rewritten…
Roadmap completion: what was done, and how far each piece was tested (2026-10-16, refreshed after rc45)	Limits that apply to every row: CI has no canvas, no GPS device, no sample point-cloud or tile data, no model/LLM in the loop for the live tests (they…
Visualization, styling and analysis gap analysis (2026-10-01)	Dated snapshot. Method: read the registry and the tool code (not the docs), counted what could be counted, and for each gap say whether it is closed i…
Sharing a workflow as a file	A workflow file is a small text file that records the steps of an analysis so a colleague can read it, point it at their own layers and run it. Collea…

Worksheets and data files

File	Kind
DPIA_SCREENING_WORKSHEET.docx	Word document
FULL_LIVE_TEST_open_issues_ROWS.csv	Result matrix or row list
FULL_LIVE_TEST_rc34_INPUTS.sha256	Checksums
FULL_LIVE_TEST_rc34_ROWS.csv	Result matrix or row list
FULL_LIVE_TEST_rc37_ROWS.csv	Result matrix or row list
GDPR_COMPLIANCE_REVIEW.docx	Word document
HAND_TEST_rc44_algorithms.csv	Result matrix or row list
HAND_TEST_rc44_tools.csv	Result matrix or row list
HAND_TEST_rc46_algorithms_plain.csv	Result matrix or row list
LIVE_TEST_rc37_README.txt	Plain-text guide
MASTER_TASK_LIST_2026-10-27.csv	Result matrix or row list
OPEN_ISSUES_COVERAGE_rc40.csv	Result matrix or row list
RC22_RESULT_MATRIX_2026-10-09.csv	Result matrix or row list
generate_persona_catalogue.py	Script
generate_tools_reference.py	Script
route_optimization_prototype.py	Script

Archive

Frozen early specifications and reviews (see the archive index).

Document	What it is
Cartogen AI — API Cost & Integration Efficiency Review	This is the single highest-leverage item: every other optimization in this document only matters if the model can actually see the tool it needs. Laye…
Auto-Reporting Recipe — Program Update Composition	A gap-analysis review proposed "schedulerecurringworkflow + generatehtmldashboard + generatesectorcoveragereport already compose into an auto-generate…
Cartogen AI — Feature List & Reality Matrix	Deliberately deferred by the user on 2026-08-10 rather than attempted (all genuinely large/roadmap-scale): real pytest-qgis harness, scheduled/recurri…
Cartogen AI — Product Requirements Document (v2)	The Cartogen AI is a QGIS plugin that lets users perform spatial analysis, data management, and cartographic output through natural-language conversat…
CVA Market-Access Recipe — What GIS Can and Can't Answer	Cash and Voucher Assistance (CVA) feasibility is a CALP Network-defined pre-condition checklist with four areas, only one of which is GIS-computable:
Destructive Tools Audit (2026-08-21)	Requested as a follow-up to docs/ENGINEERINGPRODUCTUXREVIEW2026-08-20.md §3.2's recommendation to audit which destructive tools have the preview/confi…
dock_widget.py Split: What Was Done, What Wasn't, and Why (2026-08-21)	Tracks the docs/ENGINEERINGPRODUCTUXREVIEW2026-08-20.md §3.2 recommendation to split ui/dockwidget.py (1,428 lines as of this doc, up from the review'…
Cartogen AI — Full Project Documentation	This is a single, consolidated reference for the whole project — what it is, how it's built, what's shipped versus roadmap, how it's licensed, and how…
Cartogen AI — Engineering, Product & UX Review (2026-08-20)	Requested directly by the user: a specialist read of the codebase — product engineering, software engineering, and UI/UX — with concrete recommendatio…
Cartogen AI — Humanitarian Crisis Management & Fund Allocation Feature Review	Nine new tools shipped (116 → 125), plus a provenance addition to two existing ones:
Cartogen AI — PRD Implementation Task List	Substantial functionality was built after this list was last accurate that isn't reflected in the work packages above:
JIAF Multi-Sector Composite — Specification	calculateseverityindex already implements a JIAF/INFORM-style single-sector severity index: min-max normalize indicators, weighted sum, 5-phase equal-…
Cartogen AI — License Compliance Audit	The Cartogen AI plugin source code is released under the GNU General Public License v2 (GPL-v2) or later, in compliance with the QGIS Plugin Architect…
Cartogen AI Phase 1 — Delivery Control Plan	Deliver a working AI-first humanitarian mapping application. The user describes an objective in chat; Gemini proposes a reviewable plan; approved tool…
Prompt Refinement Layer — Specification	Today, whatever the user types goes straight from ChatInputEdit into agent.run(userquery, ...) (ui/dockwidget.py → sendmessage, line ~952). Two conseq…
Archive — dated, frozen review/audit/spec/strategy docs	Everything in this folder is a dated, frozen snapshot per CONTRIBUTING.md §2's convention: each one is an accurate record of a specific review, audit,…
Route Optimization Strategy — Post-Disaster / Humanitarian Logistics	agent/tools/logisticstools.py's own module docstring is upfront about this, and it's worth repeating rather than glossing over: calculateservicearea a…
Route Risk-Scoring & No-Go Zones — Specification	Three related gaps, all confirmed against the real code before writing this:
SAM-Family Imagery Feature Extraction — Specification	CARTOGENAIPRD.md Phase 3 lists "semi-automated feature extraction from imagery" as not started. docs/SECURITYANDCOMPETITIVEREVIEW2026-08.md §B.2 found…
Cartogen AI — Status Review (2026-08-20)	Full-codebase review requested directly by the user, current as of v1.2.21. This is a snapshot, not a permanent record — treat it the way this project…
Transparency Cards	Three cards cover the actions with the largest blast radius if something goes wrong: running model-generated code, running model-generated SQL, and de…
UX & Documentation Audit — 2026-08-31	Scope: the shipped Cartogen AI Community plugin (v1.4.4) as it actually behaves and reads today — not the roadmap. Every finding below was verified by…
Cartogen AI Web Mapping Architecture — Ground-up Implementation Notes	The chat is the primary input. The map is the primary spatial output. The right panel is context and control, not a catalogue of unrelated buttons.