adds submission to gemini hackathon

Author: Josephrp

docs/configuration.md

@@ -121,17 +121,18 @@ API endpoints expect a Bearer JWT. Tokens are issued by `POST /auth/token` (subj
 
 ## LLM
 
-The orchestrator (REACT loop), judge, whitelist agent, and daily-summary cron use an LLM. Set the provider, model, and the matching API key. For **local/custom** endpoints (e.g. [Ollama](https://ollama.ai)), set `provider: custom`, `model` (e.g. `ollama/llama2` or `llama2`), and **`custom_api_base`** (e.g. `http://localhost:11434`). For **[Hugging Face Inference Providers](https://huggingface.co/docs/inference-providers)** (serverless models from Groq, Together, etc.), set `provider: huggingface`, `model` (e.g. `openai/gpt-oss-120b:groq`, `Qwen/Qwen2.5-7B-Instruct-1M`), and **`huggingface_api_key`** or `HF_TOKEN`; the client uses the HF router URL as `api_base`.
+The orchestrator (REACT loop), judge, whitelist agent, and daily-summary cron use an LLM. Set the provider, model, and the matching API key. For **local/custom** endpoints (e.g. [Ollama](https://ollama.ai)), set `provider: custom`, `model` (e.g. `ollama/llama2` or `llama2`), and **`custom_api_base`** (e.g. `http://localhost:11434`). For **[Hugging Face Inference Providers](https://huggingface.co/docs/inference-providers)** (serverless models from Groq, Together, etc.), set `provider: huggingface`, `model` (e.g. `openai/gpt-oss-120b:groq`, `Qwen/Qwen2.5-7B-Instruct-1M`), and **`huggingface_api_key`** or `HF_TOKEN`; the client uses the HF router URL as `api_base`. For **Google Gemini** (Google AI Studio), set `provider: gemini`, `model` (e.g. `gemini-2.5-flash`, `gemini-2.5-pro`), and **`gemini_api_key`** or `GEMINI_API_KEY`.
 
 | Option | Env var | Default | Description |
 |--------|---------|---------|-------------|
-| `llm.provider` | `RADIOSHAQ_LLM__PROVIDER` | `mistral` | One of: `mistral`, `openai`, `anthropic`, `custom`, `huggingface`. |
-| `llm.model` | `RADIOSHAQ_LLM__MODEL` | `mistral-large-latest` | Model name (e.g. `mistral-small-latest`, `gpt-4o`, `ollama/llama2`; for **huggingface**: `openai/gpt-oss-120b:groq`, `Qwen/Qwen2.5-7B-Instruct-1M`). |
+| `llm.provider` | `RADIOSHAQ_LLM__PROVIDER` | `mistral` | One of: `mistral`, `openai`, `anthropic`, `custom`, `huggingface`, `gemini`. |
+| `llm.model` | `RADIOSHAQ_LLM__MODEL` | `mistral-large-latest` | Model name (e.g. `mistral-small-latest`, `gpt-4o`, `ollama/llama2`; for **huggingface**: `openai/gpt-oss-120b:groq`, `Qwen/Qwen2.5-7B-Instruct-1M`; for **gemini**: `gemini-2.5-flash`, `gemini-2.5-pro`). |
 | `llm.mistral_api_key` | `RADIOSHAQ_LLM__MISTRAL_API_KEY` | `null` | Mistral API key (or set `MISTRAL_API_KEY` if your code reads it). |
 | `llm.openai_api_key` | `RADIOSHAQ_LLM__OPENAI_API_KEY` | `null` | OpenAI API key. |
 | `llm.anthropic_api_key` | `RADIOSHAQ_LLM__ANTHROPIC_API_KEY` | `null` | Anthropic API key. |
 | `llm.custom_api_base` | `RADIOSHAQ_LLM__CUSTOM_API_BASE` | `null` | **Custom provider base URL** (e.g. `http://localhost:11434` for Ollama). Passed to LiteLLM. |
 | `llm.custom_api_key` | `RADIOSHAQ_LLM__CUSTOM_API_KEY` | `null` | Custom provider API key. |
+| `llm.gemini_api_key` | `RADIOSHAQ_LLM__GEMINI_API_KEY` | `null` | **Gemini** API key (Google AI Studio; or set `GEMINI_API_KEY`). |
 | `llm.huggingface_api_key` | `RADIOSHAQ_LLM__HUGGINGFACE_API_KEY` | `null` | **Hugging Face** token for [Inference Providers](https://huggingface.co/docs/inference-providers) (or set `HF_TOKEN`). Token needs "Inference Providers" permission. |
 | `llm.huggingface_api_base` | `RADIOSHAQ_LLM__HUGGINGFACE_API_BASE` | `null` | Optional; default `https://router.huggingface.co/v1` when provider is `huggingface`. |
 | `llm.temperature` | `RADIOSHAQ_LLM__TEMPERATURE` | `0.1` | Sampling temperature (0–2). |

docs/plan-gemini-api-support.md

@@ -59,17 +59,19 @@ POSTGRES_PASSWORD=radioshaq
 # RADIOSHAQ_LLM__ANTHROPIC_API_KEY=
 # RADIOSHAQ_LLM__CUSTOM_API_BASE=
 # RADIOSHAQ_LLM__CUSTOM_API_KEY=
+# RADIOSHAQ_LLM__GEMINI_API_KEY=        # For provider: gemini (Google AI Studio)
 # RADIOSHAQ_LLM__HUGGINGFACE_API_KEY=   # For provider: huggingface (Inference Providers)
 # RADIOSHAQ_LLM__HUGGINGFACE_API_BASE=  # Optional; default https://router.huggingface.co/v1
 # RADIOSHAQ_LLM__TEMPERATURE=0.1
 # RADIOSHAQ_LLM__MAX_TOKENS=4096
 # RADIOSHAQ_LLM__TIMEOUT_SECONDS=60.0
 # RADIOSHAQ_LLM__MAX_RETRIES=3
 # RADIOSHAQ_LLM__RETRY_DELAY_SECONDS=1.0
-# Alternative: some code also reads MISTRAL_API_KEY / OPENAI_API_KEY / HF_TOKEN directly
+# Alternative: some code also reads MISTRAL_API_KEY / OPENAI_API_KEY / HF_TOKEN / GEMINI_API_KEY directly
 # MISTRAL_API_KEY=
 # OPENAI_API_KEY=
 # HF_TOKEN=   # Hugging Face token with "Inference Providers" permission (when provider is huggingface)
+# GEMINI_API_KEY=
 
 # -----------------------------------------------------------------------------
 # Memory (per-callsign memory, Hindsight, daily summaries)

docs/reference/.env.example

@@ -44,13 +44,14 @@ jwt:
 # LLM (set API key in env or here; prefer env for secrets)
 # -----------------------------------------------------------------------------
 llm:
-  provider: mistral             # mistral | openai | anthropic | custom
+  provider: mistral             # mistral | openai | anthropic | custom | huggingface | gemini
   model: mistral-large-latest
   mistral_api_key: null
   openai_api_key: null
   anthropic_api_key: null
   custom_api_base: null
   custom_api_key: null
+  gemini_api_key: null         # For provider: gemini; or set GEMINI_API_KEY
   temperature: 0.1
   max_tokens: 4096
   timeout_seconds: 60.0

docs/reference/config.example.yaml

@@ -59,6 +59,7 @@ POSTGRES_PASSWORD=radioshaq
 # RADIOSHAQ_LLM__ANTHROPIC_API_KEY=
 # RADIOSHAQ_LLM__CUSTOM_API_BASE=
 # RADIOSHAQ_LLM__CUSTOM_API_KEY=
+# RADIOSHAQ_LLM__GEMINI_API_KEY=
 # RADIOSHAQ_LLM__TEMPERATURE=0.1
 # RADIOSHAQ_LLM__MAX_TOKENS=4096
 # RADIOSHAQ_LLM__TIMEOUT_SECONDS=60.0
@@ -71,6 +72,7 @@ POSTGRES_PASSWORD=radioshaq
 # MISTRAL_API_KEY=
 # OPENAI_API_KEY=
 # HF_TOKEN=
+# GEMINI_API_KEY=
 
 # -----------------------------------------------------------------------------
 # Memory (per-callsign memory, Hindsight, daily summaries)
@@ -265,3 +267,11 @@ POSTGRES_PASSWORD=radioshaq
 # RADIOSHAQ_TTS__KOKORO_SPEED=1.0
 # ElevenLabs API key (required when provider=elevenlabs)
 # ELEVENLABS_API_KEY=
+
+# -----------------------------------------------------------------------------
+# Web UI (Vite) – used when running npm run dev or serving built assets
+# -----------------------------------------------------------------------------
+# Set in web-interface/.env or project root .env when developing the React UI.
+# VITE_RADIOSHAQ_API=http://localhost:8000
+# VITE_RADIOSHAQ_TOKEN=
+# VITE_GOOGLE_MAPS_API_KEY=   # Optional. Enables Map page, Radio field map, Transcripts "View on map". Restrict key by HTTP referrer in Google Cloud Console.

radioshaq/.env.example

@@ -44,13 +44,14 @@ jwt:
 # LLM (set API key in env or here; prefer env for secrets)
 # -----------------------------------------------------------------------------
 llm:
-  provider: mistral             # mistral | openai | anthropic | custom | huggingface
+  provider: mistral             # mistral | openai | anthropic | custom | huggingface | gemini
   model: mistral-large-latest
   mistral_api_key: null
   openai_api_key: null
   anthropic_api_key: null
   custom_api_base: null
   custom_api_key: null
+  gemini_api_key: null         # For provider: gemini; or set GEMINI_API_KEY
   huggingface_api_key: null     # For provider: huggingface; or set HF_TOKEN
   huggingface_api_base: null   # Optional; default https://router.huggingface.co/v1
   temperature: 0.1

radioshaq/config.example.yaml

@@ -0,0 +1,42 @@
+# Demo environment profiles
+
+Summary of **environment variables** for running the Live HackRF + LLM demo suite. For full WSL/HackRF setup and Option C env, see [scripts/demo/demo-hackrf-full.md](../scripts/demo/demo-hackrf-full.md).
+
+**Live demos use real hardware and real LLM:** HackRF RX/TX and LLM providers are not stubbed in the documented demo flows. Set the env below and attach a HackRF; use `--require-hardware` in the relevant demo scripts to fail fast if SDR TX is not configured.
+
+## Agent and API hooks (how demos drive the system)
+
+- **radio_tx (RadioTransmissionAgent):** Invoked via `POST /radio/send-audio` (multipart WAV) and `POST /radio/send-tts` (JSON body with `message`, optional `frequency_hz`, `mode`). Requires `radio.sdr_tx_enabled=true` and `radio.sdr_tx_backend=hackrf` for HackRF; when not set or no hardware, the TX agent may still run and return success/false (e.g. "Rig manager not configured"). Compliance checks run before TX.
+- **radio_rx_audio (RadioAudioReceptionAgent):** No one-off "start monitor" HTTP endpoint. The **voice listener** (server lifespan) starts the agent when `radio.audio_input_enabled` and `radio.voice_listener_enabled` (or `audio_monitoring_enabled`) are true. Demos that need voice RX either run HQ with that config and poll `GET /api/v1/audio/pending` and `GET /transcripts`, or use `POST /messages/from-audio` to simulate inbound voice.
+- **radio_rx (RadioReceptionAgent):** Used by the band listener (injection queue consumer) or by tasks submitted via the orchestrator. Demos inject via `POST /inject/message` or `POST /messages/inject-and-store`; the band listener (when enabled) or a process-driven task consumes from the queue.
+- **Orchestrator / Judge:** `POST /messages/process` (body: `message` or `text`, optional `channel`, `chat_id`, `sender_id`) runs the REACT loop and routes to agents. Used by run_orchestrator_judge_demo and run_scheduler_demo.
+- **WhitelistAgent:** Invoked via `POST /messages/whitelist-request` (JSON or multipart with audio). Orchestrator evaluates and may call the whitelist agent; result in response or completed_tasks.
+- **SchedulerAgent:** No direct HTTP endpoint; reached when the orchestrator selects it for a scheduling request (e.g. "Schedule a call for X with Y at Z"). Requires DB with coordination_events for persistence.
+
+## HQ process (`uv run radioshaq run-api`)
+
+- **Mode + JWT:** `RADIOSHAQ_MODE=hq`, `RADIOSHAQ_JWT__SECRET_KEY` (must match receiver `JWT_SECRET`).
+- **Receiver uploads:** `RADIOSHAQ_RADIO__RECEIVER_UPLOAD_STORE=true`, `RADIOSHAQ_RADIO__RECEIVER_UPLOAD_INJECT=true`.
+- **HackRF SDR TX:** `RADIOSHAQ_RADIO__SDR_TX_ENABLED=true`, `RADIOSHAQ_RADIO__SDR_TX_BACKEND=hackrf`.
+- **Message bus consumer:** `RADIOSHAQ_BUS_CONSUMER_ENABLED=1`.
+- **LLM:** e.g. `RADIOSHAQ_LLM__PROVIDER=mistral`, `MISTRAL_API_KEY` / `RADIOSHAQ_LLM__MISTRAL_API_KEY`.
+- **ASR/TTS:** e.g. `ELEVENLABS_API_KEY`, `RADIOSHAQ_TTS__PROVIDER=elevenlabs`.
+- **Voice listener (for voice_rx_audio demos):** `RADIOSHAQ_RADIO__AUDIO_INPUT_ENABLED=true`, `RADIOSHAQ_RADIO__VOICE_LISTENER_ENABLED=true`, `RADIOSHAQ_RADIO__DEFAULT_BAND=2m`.
+- **Twilio:** Omit or leave unset for no-Twilio demos; set for Option C with SMS/WhatsApp.
+
+## Remote receiver process (`uv run radioshaq run-receiver`)
+
+- **JWT:** `JWT_SECRET` = same as HQ `RADIOSHAQ_JWT__SECRET_KEY`.
+- **Identity:** `STATION_ID=HACKRF-DEMO`.
+- **HackRF:** `SDR_TYPE=hackrf`, `HACKRF_INDEX=0`.
+- **HQ upload:** `HQ_URL=http://localhost:8000`, `HQ_TOKEN=<from POST /auth/token>`.
+- **Demod:** `RECEIVER_MODE=nfm`, `RECEIVER_AUDIO_RATE=48000`.
+
+## Demo scripts
+
+- **Base URL:** Pass `--base-url http://localhost:8000` (or remote). Scripts obtain a JWT via `POST /auth/token` (subject/role/station_id).
+- **Extras:** `uv sync --extra hackrf` (receiver + stream), `uv sync --extra voice_tx` (HackRF TX from HQ), `uv sync --extra audio` (ASR) as needed.
+
+## Database
+
+- **Postgres:** `RADIOSHAQ_DATABASE__POSTGRES_URL` or default (e.g. Docker on 5434). Run `uv run radioshaq launch docker` then `cd radioshaq && uv run alembic upgrade head` before demos that use transcripts or registry.

radioshaq/docs/demo-env-profiles.md

@@ -152,7 +152,7 @@ services:
       - HINDSIGHT_API_LLM_PROVIDER=${RADIOSHAQ_LLM__PROVIDER:-${HINDSIGHT_API_LLM_PROVIDER:-openai}}
       - HINDSIGHT_API_LLM_MODEL=${RADIOSHAQ_LLM__MODEL:-${HINDSIGHT_API_LLM_MODEL:-gpt-4o-mini}}
       # API key: first non-empty of RadioShaq keys, then generic keys
-      - HINDSIGHT_API_LLM_API_KEY=${RADIOSHAQ_LLM__OPENAI_API_KEY:-${OPENAI_API_KEY:-${RADIOSHAQ_LLM__MISTRAL_API_KEY:-${MISTRAL_API_KEY:-${RADIOSHAQ_LLM__ANTHROPIC_API_KEY:-${ANTHROPIC_API_KEY:-${RADIOSHAQ_LLM__CUSTOM_API_KEY:-${HINDSIGHT_API_LLM_API_KEY:-}}}}}}}}
+      - HINDSIGHT_API_LLM_API_KEY=${RADIOSHAQ_LLM__OPENAI_API_KEY:-${OPENAI_API_KEY:-${RADIOSHAQ_LLM__MISTRAL_API_KEY:-${MISTRAL_API_KEY:-${RADIOSHAQ_LLM__ANTHROPIC_API_KEY:-${ANTHROPIC_API_KEY:-${RADIOSHAQ_LLM__GEMINI_API_KEY:-${GEMINI_API_KEY:-${RADIOSHAQ_LLM__CUSTOM_API_KEY:-${HINDSIGHT_API_LLM_API_KEY:-}}}}}}}}}}
       # Custom base URL (e.g. OpenAI-compatible or Mistral endpoint)
       - HINDSIGHT_API_LLM_BASE_URL=${RADIOSHAQ_LLM__CUSTOM_API_BASE:-${HINDSIGHT_API_LLM_BASE_URL:-}}
       # Same Postgres as RadioShaq (postgres service, db radioshaq; pgvector in postgres/init/02-pgvector.sql)

radioshaq/infrastructure/local/docker-compose.yml

@@ -24,6 +24,7 @@
     "anthropic_api_key",
     "custom_api_key",
     "huggingface_api_key",
+    "gemini_api_key",
 }
 
 

radioshaq/radioshaq/api/routes/config_routes.py

@@ -158,10 +158,15 @@ async def get_operators_nearby(
         recent_only=recent_hours > 0,
         recent_hours=recent_hours,
     )
+    # Ensure each operator has last_seen_at for mapping clients (alias of timestamp)
+    operators_for_response = [
+        {**op, "last_seen_at": op.get("last_seen_at") or op.get("timestamp")}
+        for op in operators
+    ]
     return {
         "latitude": latitude,
         "longitude": longitude,
         "radius_meters": radius_meters,
-        "operators": operators,
-        "count": len(operators),
+        "operators": operators_for_response,
+        "count": len(operators_for_response),
     }