Skip to content

160-tool MCP server for real-time global intelligence from free public APIs: markets, liquidity, prediction markets, SEC filings, conflict, military, aircraft-cohort early warning, migration and border flows, cyber, climate, news and more. Dated observation history, 29 research prompts, cited briefs, geofences, SSE dashboard. MIT, no paid API keys.

Release 0.4.0: fold the data-honesty fixes into the release entry

Phoenix Agent (Marcs-Mac-Studio)a08efa5f651e
diff --git a/CHANGELOG.md b/CHANGELOG.md
index 47f2fb9..3fbf847 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -1,34 +1,10 @@
# Changelog
-## Unreleased
-
-The data-honesty follow-up: the three verified fail-quietly bugs the
-0.4.0 test waves documented are now fixed, and coverage is gated in CI.
-
-### Fixed
-- A UNHCR outage no longer reads as zero refugees worldwide:
- `intel_displacement_summary`'s failure shape carries `error`,
- `degraded`, and `reason: unhcr_fetch_failed` instead of bare zeroed
- totals (#22).
-- Climate zones whose fetch failed are named in `unavailable_zones`
- (with `degraded: true`) instead of silently vanishing; a full
- Open-Meteo outage is an `error` with a reason, and an invalid zone
- filter now says so and lists the valid keys (#23).
-- "Resolved: Major outage" post-mortems no longer count as active
- critical incidents (resolution status outranks incident keywords),
- and a provider whose status feed is unreachable is named in
- `unavailable_providers` instead of masquerading as healthy with no
- incidents (#24).
-
-### Added
-- CI coverage ratchet: `--cov-fail-under=80` on the test job
- (measured 81% on 2026-09-01; floor set one point under for platform
- variance). The floor only moves up.
-
## 0.4.0 - 2026-09-01
-Geofences that survive the dateline, notice change, and a test suite
-that reaches the layers the old one never imported.
+Geofences that survive the dateline, notice change, a test suite that
+reaches the layers the old one never imported, and outages that say so
+instead of reading as good news.
### Added
- `intel_aoi_changes` (+1 tool): geofence change detection — what
@@ -58,7 +34,11 @@ that reaches the layers the old one never imported.
`sources/` modules (intelligence, cyber, climate, displacement,
fleet, prediction, service_status, maritime, military parsing) had
0-30% statement coverage; measured overall coverage was 59% before
- this release despite a green 309-test suite.
+ this release despite a green 309-test suite. At release: 597
+ non-smoke tests, 81% statement coverage.
+- CI coverage ratchet: `--cov-fail-under=80` on the test job
+ (measured 81% on 2026-09-01; floor set one point under for platform
+ variance). The floor only moves up.
### Fixed
- Antimeridian AOIs: the bounding box derived from an AOI's radius was
@@ -90,6 +70,19 @@ that reaches the layers the old one never imported.
- `__version__` no longer drifts from `pyproject.toml` (it had been
stuck at 0.1.0 since the first release): it now derives from the
installed package metadata at runtime.
+- A UNHCR outage no longer reads as zero refugees worldwide:
+ `intel_displacement_summary`'s failure shape carries `error`,
+ `degraded`, and `reason: unhcr_fetch_failed` instead of bare zeroed
+ totals (#22).
+- Climate zones whose fetch failed are named in `unavailable_zones`
+ (with `degraded: true`) instead of silently vanishing; a full
+ Open-Meteo outage is an `error` with a reason, and an invalid zone
+ filter now says so and lists the valid keys (#23).
+- "Resolved: Major outage" post-mortems no longer count as active
+ critical incidents (resolution status outranks incident keywords),
+ and a provider whose status feed is unreachable is named in
+ `unavailable_providers` instead of masquerading as healthy with no
+ incidents (#24).
## 0.3.0 - 2026-08-16
diff --git a/README.md b/README.md
index 3f8efef..6e8acd7 100644
--- a/README.md
+++ b/README.md
@@ -571,7 +571,7 @@ Everything else uses free, unauthenticated public APIs.
```bash
pip install -e ".[dev]"
-pytest # 595 tests (613 total, 18 live-network smoke tests deselected by default)
+pytest # 597 tests (615 total, 18 live-network smoke tests deselected by default)
pytest --cov=world_intel_mcp # with coverage
pytest tests/test_forex.py -v # single module
```
diff --git a/ROADMAP.md b/ROADMAP.md
index 0d711f5..3ad91a6 100644
--- a/ROADMAP.md
+++ b/ROADMAP.md
@@ -23,7 +23,7 @@
| MCP tool parity | 122 tools declared in `TOOLS`; 122 routed in `_dispatch()` | :white_check_mark: | Keep as an invariant |
| Optional vector runtime | Missing `qdrant-client` / `fastembed` previously surfaced as runtime failures | :white_check_mark: Fixed | Vector features now degrade cleanly and report availability |
| Base-environment test run | `pytest -q` fails collection without dev extras because `respx` is not installed | :yellow_circle: | Run `pip install -e ".[dev]"` before full-suite validation |
-| Test coverage truth | Was 59% overall on 2026-09-01 morning (`server.py`/`cli.py`/`collector.py` at 0%, `analysis/` NLP modules 0-14%, `sources/intelligence.py` 22%). After the same-day test waves: **81% overall, 595 tests** (analysis modules 96-100%, `sources/intelligence.py` 93%, `server.py` 47% via import-based registry tests). Remaining zeros: `cli.py` (1,076 stmts); `collector.py` at 20% (map verified, run loop untested) | :yellow_circle: Improved | CI coverage gate that ratchets (Phase 24); `cli.py` smoke tests |
+| Test coverage truth | Was 59% overall on 2026-09-01 morning (`server.py`/`cli.py`/`collector.py` at 0%, `analysis/` NLP modules 0-14%, `sources/intelligence.py` 22%). After the same-day test waves: **81% overall, 597 tests** (analysis modules 96-100%, `sources/intelligence.py` 93%, `server.py` 47% via import-based registry tests). Remaining zeros: `cli.py` (1,076 stmts); `collector.py` at 20% (map verified, run loop untested) | :yellow_circle: Improved | CI coverage gate that ratchets (Phase 24); `cli.py` smoke tests |
| Geofence correctness | Antimeridian AOIs lost the far side of the dateline (bbox clamp); pipelines/cables matched on endpoints only | :white_check_mark: Fixed in Phase 22 | Segment distance + split bboxes shipped with tests |
| Security posture | Issue #21 (external report) assessed; no shell/eval/exec, SQL parameterized throughout, report paths server-generated | :white_check_mark: | SECURITY.md added with explicit threat model; cache db now 0600 |
| Documentation drift | Prior roadmap documented 89/110 tools while the codebase now exposes 113 | :white_check_mark: Updated below | Keep roadmap synced with phase increments |
@@ -595,7 +595,7 @@ import-based tests make this refactor safe to attempt).
| Tool parity | 122 / 122 | `TOOLS` and `_dispatch()` are aligned (now machine-checked by import-based tests, not text scans) |
| Static datasets | 18 | Bases, ports, pipelines, nuclear, cables, datacenters, spaceports, minerals, exchanges, trade routes, cloud regions, financial centers |
| RSS feeds | 119 | 24 categories |
-| Tests in repo | 613 | 595 non-smoke tests + 18 live smoke tests (measured 2026-09-01: `pytest -q` -> 595 passed, 18 deselected); full suite requires `.[dev]` |
+| Tests in repo | 615 | 597 non-smoke tests + 18 live smoke tests (measured 2026-09-01: `pytest -q` -> 597 passed, 18 deselected); full suite requires `.[dev]` |
| Statement coverage | 81% | Measured 2026-09-01 full-suite `--cov`; was 59% that morning. Biggest remaining zeros: `cli.py` (0%), `collector.py` run loop (20%) |
| Primary remaining gaps | `cli.py` tests, coverage gate, `server.py` refactor | See Planned Phases 23-26 |