chore(deps): Bump the bowire group with 1 update

[dependabot]

Updated Kuestenlogik.Bowire from 1.8.0 to 1.9.0.

Release notes

Sourced from Kuestenlogik.Bowire's releases.

1.9.0

Highlights — Tier 4 of the security roadmap lands

Bowire's product positioning has been "Burp Suite for the non-HTTP protocols, with schema-awareness, self-hosted, with AI-assisted threat modeling via MCP." v1.9 closes the AI-assisted half of that promise — every discovered API surface in the workbench can now be analyzed by an LLM that grounds itself in the actual schema + traffic, not generic web knowledge.

• AI threat-model (#​59) — rank every discovered endpoint by attack-surface risk. The model reads the API's input/output shapes, auth posture, and recent traffic, and returns a sorted list with reasoning per endpoint. Sets the scan order for everything that follows.
• AI Nuclei-template suggestion (#​60) — per endpoint, generate Nuclei templates targeting the identified risks. Bowire feeds the model the OpenAPI / proto / GraphQL schema; the model returns YAML that drops straight into the Nuclei runner.
• AI findings triage (#​61) — every Nuclei hit gets a real-vs-false-positive verdict + a concrete fix suggestion. Cuts triage time per finding by an order of magnitude on a noisy scan.
• AI schema-aware fuzz values (#​62) — boundary inputs per field, derived from the schema's types + constraints. Composes with the existing fuzz harness.
• AI Settings UI (#​63) — pick the provider (Ollama, LM Studio, BYOK cloud) + endpoint + model directly from the workbench's Settings dialog. No appsettings.json edits, no restart.

The roadmap calls this Tier 4. Tiers 1–3 (record-as-attack, fuzz UI, MITM proxy) shipped through v1.4.x; Tier 5 (auto-exploit / proof-of-vulnerability) is its own framing and a different risk posture, deliberately out of scope here.

Also in this release

The 94 commits since v1.8.0 carry a lot more than the marquee. Highlights of the supporting work:

Security as a first-class drawer, with a tier that doesn't need AI (#​111 + #​112)

• Separate Security drawer. Threat-Model + Nuclei template suggestion left the AI drawer for their own surface (shield icon in the topbar, Ctrl/Cmd+Shift+S). The AI drawer stays focused on conversational assistance (hints + chat). Mental model: AI = assistant, Security = scanner / analysis.
• Heuristic ranking tier — no AI required. Default ranking is now a deterministic rule engine (ThreatHeuristic in core): verb-based mutation scoring, BOLA-pattern detection on path params, admin / auth / PII path matching, anonymous-auth bumps, sensitive-field-name detection in the input shape. Sub-millisecond per endpoint. Each ranked row carries a ruleTrace so users can audit which rules fired against which endpoint.
• AI tier becomes opt-in. A tier toggle in the drawer lets the operator switch to AI-assisted ranking when they want semantic adjustments on top of the heuristic. Default heuristic → security tooling works on installs that don't have or don't want AI.

AI assistant — full MCP-style tool calling (#​89, #​108, #​109)

• Phase 1: chat grounding. Every chat send prepends a workbench-state snapshot (loaded URLs + service names + selected method's full schema + recent calls) as a system prompt. The model answers from real data instead of generic web knowledge.
• Phase 2: read-only tools (#​108). Three Microsoft.Extensions.AI AIFunctions available on every chat request: bowire_list_services, bowire_describe_method, bowire_recent_history. The model calls them mid-conversation to drill in. The Ollama path is wrapped in FunctionInvokingChatClient so the tool-call loop actually round-trips. Tool calls render as visible "Consulted X" steps in the chat transcript with collapsible args.
• Phase 3: invoke tool (#​109). Opt-in via the "Allow AI to invoke methods" toggle in the drawer header. When OFF (default), bowire_invoke isn't even registered — the model literally cannot try. When ON, the AI dispatches through protocol.InvokeAsync (same path /api/invoke uses) and every call writes a JSONL row to ~/.bowire/.ai-actions.jsonl for audit. Toggle is session-only — never persisted to localStorage.
• Phase 4: UI navigation (#​109). bowire_open_method tool lets the AI navigate the workbench to the right method ("Show me pet.findPetsByStatus" → workbench opens the request pane on it). Always available, no side effects beyond UI state.
• "Thinking…" feedback in the drawer. Local models with tool calling take 15-45 s per turn. The chat now shows a live (N s) counter with a pulsing accent dot + a Cancel button that aborts the fetch — no more wondering whether the request hung.

Workbench UX

• AI side panel moves into a right-side drawer (#​90) with a topbar toggle + Ctrl/Cmd+Shift+A shortcut. Persists across method/service switches. Was a peer of Response/Logs/Code in the response-pane tab strip, which forced you to choose between seeing the AI and seeing the response.
• Connection-state pill in the topbar (#​93). Aggregate state for every configured discovery URL — green when all connected, amber when partial / connecting, red when any failed. Hover for a per-URL breakdown with service counts + the failure message.
• Body sub-tabs (#​85). GraphQL Body now splits Query / Variables / Selection-set into a sub-tab strip; REST / gRPC / JSON-RPC get protocol-aware Form / JSON pair labels (gRPC: Message / JSON; JSON-RPC: Params / JSON; REST: Form / Body). Was a vertical stack of three surfaces; now one surface at a time.
• Filter services by discovery URL in the sidebar's filter popup. Multi-URL setups can narrow to one origin.
• Theme toggle dropped its visible label — tooltip still carries the state + next action; the topbar right cluster reads cleaner.

API conventions

• RFC 7807 ProblemDetails (#​88) is the new shape for every API error response. application/problem+json with stable type URNs (urn:bowire:ai:model-not-found, urn:bowire:discovery:no-match, …), structured detail, and typed extensions per error class. Backward-compatible: every body still carries an error field set to the title so legacy readers degrade gracefully.
• AI 404 errors now actionable (#​87). Missing model surfaces as "Model 'X' isn't available — pull it with ollama pull X or pick a different model in Settings → AI", with a links: [{rel: configure}] extension pointing back at the settings page.

Bug fixes

• #​65 — URL input in the sidebar was read-only despite lockServerUrl=false. el() helper coerced undefined attribute values to the literal string "undefined", which the browser treats as truthy.
• #​66 — Settings → Plugins tab was in a fetch/render infinite loop; the plugins-tab fetches now fire once per tab-open instead of once per render.
• #​81 — Standalone CLI without --url was falling back to embedded UI mode because the heuristic checked serverUrls.length instead of trusting config.embeddedMode.
• #​82 — Standalone CLI's discovery endpoint short-circuited to [] even when a runtime URL was passed via the sidebar.
• #​83 — Plugin discovery probes ran sequentially (12 plugins × ~2-3 s ≈ 30 s) and the frontend timed out at 12 s. Now parallel with Task.WhenAll + a 8 s per-probe ceiling.
• #​84 — Standalone discovery without a URL fell back to the workbench's own URL, which the JSON-RPC plugin matched with a phantom "Methods" stub.
• #​86 — gRPC/HTTP transcoding toggle rendered on every REST method because the predicate didn't check method.source === 'grpc'.

Docs

• New Topbar UI Guide page documenting the brand + command palette + connection pill + env selector + AI drawer.
  ... (truncated)

Commits viewable in compare view.

Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting @dependabot rebase.

---

Dependabot commands and options

You can trigger Dependabot actions by commenting on this PR:

• @dependabot rebase will rebase this PR
• @dependabot recreate will recreate this PR, overwriting any edits that have been made to it
• @dependabot show <dependency name> ignore conditions will show all of the ignore conditions of the specified dependency
• @dependabot ignore <dependency name> major version will close this group update PR and stop Dependabot creating any more for the specific dependency's major version (unless you unignore this specific dependency's major version or upgrade to it yourself)
• @dependabot ignore <dependency name> minor version will close this group update PR and stop Dependabot creating any more for the specific dependency's minor version (unless you unignore this specific dependency's minor version or upgrade to it yourself)
• @dependabot ignore <dependency name> will close this group update PR and stop Dependabot creating any more for the specific dependency (unless you unignore this specific dependency or upgrade to it yourself)
• @dependabot unignore <dependency name> will remove all of the ignore conditions of the specified dependency
• @dependabot unignore <dependency name> <ignore condition> will remove the ignore condition of the specified dependency and ignore conditions

[dependabot]

Labels

The following labels could not be found: dependencies, nuget. Please create them before Dependabot can add them to a pull request.

Please fix the above issues or remove invalid values from dependabot.yml.