AquaLens pulls recent Sentinel-2 imagery, computes six water-quality indices, fuses optional field evidence, and writes a grounded risk brief — so the field team knows where to sample first.
Risk
Medium · 0.46
Routine urgency
Spectral indices
Each index isolates a different water-quality signal. AquaLens computes them on the same Sentinel-2 acquisition and aggregates over the water mask.
formula · (NIR − SWIR) / (NIR + SWIR)
Tells us where water actually is in the scene.
How it works
The deterministic numeric core runs first and produces the auditable numbers. The Gemini agent layer wraps it: agents choose inputs and write prose, but they can't move the risk band. Every decision lands in the per-session trace.
Step 01
search, paste coordinates, or tap the map
AquaLens turns the chosen point into a ~1 km buffer polygon and records a date window. A water-mask check classifies the AOI as water, mixed, or land.
Step 02
Microsoft Planetary Computer · STAC query
The backend searches the Sentinel-2 L2A archive for the freshest scene that intersects the AOI under the cloud-cover ceiling, signs the COG asset URLs, and reads only the bands it needs.
Step 03
deterministic numpy + audit-friendly weights
Six band-math indices (NDWI · MNDWI · NDTI · NDCI · NDVI · WRI) are computed over the water mask, with full provenance for every value.
Step 04
weighted, unit-tested, reproducible
A deterministic weighted model turns the index aggregates into a 0–100 score, a level (low / medium / high), and an urgency tier. The agent layer is never allowed to move these numbers.
Agent 1
GeminiGemini thinking mode
Reads the AOI plus its history and decides per-agent budgets. Always schedules Scout, Analyst, and Reporter; schedules Historian only when prior sessions exist.
Agent 2
GeminiFunction calling + Gemini Vision
Evaluates candidate scenes and asks Gemini Vision to look at the actual RGB thumbnail. If haze sits over the AOI it re-queries with a tighter cloud bound.
Agent 3
GeminiHistory + Search grounding + memory
Pulls prior sessions, runs a Mann-Kendall significance test in Gemini's Python sandbox, cites real local news via Google Search grounding + URL Context, and writes a distilled note back to per-water-body memory for next time.
Agent 4
GeminiStructured output + critique loop
Drafts recommendation + reasoning + limitations against the deterministic numbers. A separate Gemini call critiques the draft against the hard rules; the Analyst rewrites once if anything fails. Both drafts land in the trace.
Agent 5
GeminiStructured response schema
Turns the deterministic numbers plus Scout / Historian / Analyst outputs into the public-facing card: likely-safe · use caution · avoid, guidance for adults and for pets and kids, explicit limitations, and citations when grounded context exists. Falls back to a deterministic summary if the Gemini call fails.
Every agent run is captured in the Agentic workflow card on the session page. If any agent fails the orchestrator records the failure and falls back to a deterministic path, so a session always produces a usable brief.
Agent surface
The deterministic numeric core is wrapped by a small graph of specialist agents. Each one has a focused job, a constrained tool surface, and a structured-output contract — and every call is captured in a per-session trace.
Agent 1
Reads the AOI plus its history and decides which specialists to invoke and what their tool / time budgets are. Always schedules Scout, Analyst, and Reporter for water AOIs; schedules Historian when prior sessions exist. Failed plan? The orchestrator falls back to a baseline schedule so the session still completes.
Agent 2
Function calling + multimodal vision
Calls Planetary Computer for Sentinel-2 candidates, then asks Gemini Vision to look at the actual RGB thumbnail. If haze sits over the AOI it re-queries with a tighter cloud bound.
Tools
Agent 3
Code execution + Search grounding + memory
Pulls prior sessions, runs a Mann-Kendall significance test in Gemini's Python sandbox, cites real news via Google Search grounding + URL Context, and writes a distilled note back to per-water-body memory for next time.
Tools
Agent 4
Structured output + critique loop
Drafts recommendation + reasoning + limitations against the deterministic numbers. A separate Gemini call critiques the draft against the hard rules; the Analyst rewrites once if anything fails. Both drafts land in the trace.
Tools
Agent 5
Structured response schema
Turns Scout / Historian / Analyst outputs plus the deterministic risk numbers into the public-facing card: tone (likely-safe / use caution / avoid), guidance for adults and for pets and kids, explicit limitations, and citations when grounded context exists. Falls back to a deterministic citizen summary if the Gemini call fails.
Tools
agent_traces
JSONB log per session
Every tool call, argument, result, error, and token count — surfaced live in the Agentic workflow card.
agent_memory
vectors per water body
text-embedding-004 vectors with a pgvector(768) HNSW cosine index. Next session for the same lake recalls semantically related notes.
Gemini capabilities exercised end-to-end
Built on open data and free infrastructure
Ready when you are
Drawing the polygon takes ten seconds. The first scene, indices, and risk brief land less than a minute later — no account required.