feat: add self-hosted SearXNG research mode

docs/DEPLOYMENT.md

@@ -17,8 +17,7 @@ After the first successful login, remove both values from `.env`, restart the AP
 
 Copy `.env.example` to an untracked deployment environment file. Production requires a secret-manager supplied `SESSION_SECRET` of at least 32 characters and refuses `AUTOMATED_OUTREACH_ENABLED=true`. Keep bootstrap credentials one-time only; remove and rotate them after provisioning. Never place secrets in images, Compose YAML, logs, backups, or public web roots.
 
-For optional native Nous Portal Chat Completions tool-calling discovery, set these
-server-side variables:
+For criteria-first AI web research without a paid scraper, use the self-hosted mode:
 
 ```dotenv
 AI_RESEARCH_PROVIDER=nous_portal
@@ -26,20 +25,24 @@ NOUS_API_KEY=<Nous Portal API key>
 NOUS_MODEL=Hermes-4-405B
 NOUS_BASE_URL=https://inference-api.nousresearch.com/v1
 NOUS_ALLOWED_HOSTS=inference-api.nousresearch.com
-FIRECRAWL_API_KEY=<Firecrawl API key>
-FIRECRAWL_BASE_URL=https://api.firecrawl.dev/v2
-FIRECRAWL_ALLOWED_HOSTS=api.firecrawl.dev
+SEARXNG_BASE_URL=http://searxng:8080
+SEARXNG_ALLOWED_HOSTS=searxng
+SEARXNG_SECRET_KEY=<random secret>
 ```
 
-The adapter calls Nous at `/chat/completions` with strict `web_search` and
-`scrape_website` function tools. Tool calls are executed only against the
-allowlisted Firecrawl-compatible API, capped at 4 calls and 16 KiB per tool
-result. Prompt-injection-shaped criteria are rejected and tool/page content is
-untrusted data. The final model response is parsed only as structured JSON
-HTTPS targets; the existing crawler performs SSRF validation and persists
-fetched-page evidence. Missing either key, unavailable providers, unsafe base
-URLs, malformed tool calls, oversized responses, and exhausted budgets fail
-closed. Status metadata never includes secrets.
+Compose starts SearXNG without publishing a host port. The API reaches it only
+on the internal Compose network; SearXNG alone has a separate egress network for
+its upstream search engines. `web_search` calls SearXNG's JSON endpoint and
+`scrape_website` uses the API's bounded SSRF-safe scanner. No Firecrawl key is
+required. Firecrawl variables remain an optional legacy compatibility fallback.
+
+The Nous adapter calls `/chat/completions` with strict `web_search` and
+`scrape_website` function tools. Tool calls are capped at 4 and each result at
+16 KiB; criteria, model output, page text, redirects, and candidate URLs remain
+bounded and untrusted. Only structured HTTPS targets are accepted and the
+existing crawler fetches/persists evidence. Missing/unsafe configuration,
+unavailable providers, malformed calls, oversized responses, SSRF targets, and
+exhausted budgets fail closed. Status metadata never includes secrets.
 
 The prior OpenAI Responses and generic provider variables remain supported only
 as compatibility adapters.