@lazyingart/agintiflow 0.17.6 → 0.19.0

package/README.md

@@ -46,15 +46,17 @@ aginti --list-profiles
 aginti --sandbox-status
 ```
 
-On first interactive use, if no main model key is detected, `aginti` opens an auth wizard. Use Up/Down to choose DeepSeek, OpenAI, or Qwen, paste the key, and press Enter to save it to the project-local ignored file `.aginti/.env` with `0600` permissions. The wizard points to DeepSeek keys at `https://platform.deepseek.com/api_keys` and OpenAI keys at `https://platform.openai.com/api-keys`. It then offers the optional auxiliary image key; press Esc to skip. You can rerun it even when keys already exist:
+On first interactive use, if no main model key is detected, `aginti` opens an auth wizard. Use Up/Down to choose DeepSeek, OpenAI, Qwen, or Venice, paste the key, and press Enter to save it to the project-local ignored file `.aginti/.env` with `0600` permissions. The wizard points to DeepSeek keys at `https://platform.deepseek.com/api_keys`, OpenAI keys at `https://platform.openai.com/api-keys`, and Venice at `https://venice.ai`. It then offers the optional auxiliary image key; press Esc to skip. You can rerun it even when keys already exist:
 
 ```bash
 aginti auth
 aginti auth openai
+aginti auth venice
 # inside chat, use /login or /auth for the same wizard
 # or non-interactively:
 printf '%s' "$DEEPSEEK_API_KEY" | aginti keys set deepseek --stdin
 printf '%s' "$QWEN_API_KEY" | aginti keys set qwen --stdin
+printf '%s' "$VENICE_API_KEY" | aginti keys set venice --stdin
 
 # optional image-generation auxiliary skill:
 aginti login grsai
@@ -75,6 +77,8 @@ Inside chat, type normal requests such as `write a small Python CLI app with tes
 
 For code edits, AgInTiFlow routes patch/refactor/database-style tasks to DeepSeek v4 pro by default and exposes `apply_patch` as a deterministic workspace tool. It supports exact replacements, Codex-style patch envelopes, and unified diffs, with preflight checks, path guardrails, hashes, and compact per-file diffs. See [docs/patch-tools.md](docs/patch-tools.md).
 
+Model choice is role-based: `/route` defaults to DeepSeek V4 Flash, `/model` or `/main` defaults to DeepSeek V4 Pro, `/spare` defaults to OpenAI GPT-5.4 medium, `/wrapper` defaults to Codex GPT-5.5 medium when enabled, and `/auxilliary` defaults to GRS AI/Nano Banana for image tools. See [docs/model-selection.md](docs/model-selection.md) and run `aginti models`.
+
 For larger repositories, use `--profile large-codebase` or choose **Large codebase engineering** in the web UI. The web default stays **Auto**, and Auto now escalates codebase/system/debugging prompts to the same engineering loop when needed. Complex work routes to DeepSeek v4 pro, starts with `inspect_project`, then uses search/read/patch/check loops inspired by Codex, Copilot SDK, Claude Code, Gemini CLI, Qwen, and Claw Code. See [docs/large-codebase-engineering.md](docs/large-codebase-engineering.md).
 
 AgInTiFlow can also spend cheap DeepSeek calls on parallel scout notes before the main executor starts a complicated task. It first writes a bounded project map to `.aginti/codebase-map.json`, then runs scouts for architecture, implementation, review, research, context mapping, tests, git workflow, integration, symbol tracing, and dependency risks. A coordinator Swarm Board is injected for the main agent and saved as `artifacts/scout-blackboard.json` in the session. The executor still does the real file/shell/browser work itself. Disable with `--no-parallel-scouts` or set `--scout-count 1..10`.
@@ -83,7 +87,7 @@ The next productive-agent roadmap is tracked in [docs/productive-agent-roadmap.m
 
 For current docs, install errors, package/toolchain setup, and source discovery, the agent has a guarded `web_search` tool. It returns compact search results without browser search-engine loops and respects configured domain allowlists. Disable with `--no-web-search`.
 
-For raster image work, AgInTiFlow has an optional `image_generation` skill backed by the `generate_image` tool and a local `GRSAI` key. The skill tells DeepSeek when image generation is appropriate; the tool calls GRS AI Nano Banana, saves manifests/images under `artifacts/images`, and sends the result to the canvas. See [docs/auxiliary-image-generation.md](docs/auxiliary-image-generation.md).
+For raster image work, AgInTiFlow has optional `image_generation` and `venice_image_generation` skills backed by the `generate_image` tool and local `GRSAI` or `VENICE_API_KEY` credentials. The skill tells the model when image generation is appropriate; the tool saves manifests/images under `artifacts/images` and sends the result to the canvas. See [docs/auxiliary-image-generation.md](docs/auxiliary-image-generation.md) and [references/venice-model-reference.md](references/venice-model-reference.md).
 
 For long-running shell work, AgInTiFlow exposes host-side tmux tools when the shell tool is enabled: `tmux_list_sessions`, `tmux_capture_pane`, `tmux_send_keys`, and `tmux_start_session`. Use normal prompts such as `start this test server in tmux and monitor it` or `check my tmux session`. The agent captures panes before interacting, redacts outputs, blocks secret-like sends, and avoids sending sudo passwords or destructive commands. Tmux is intentionally host-side: Docker `run_command` containers are short-lived, so tmux servers started there cannot persist.
 
@@ -273,7 +277,7 @@ Defaults:
 | `smart` | DeepSeek | Fast for normal tasks, pro for complex tasks | `AGENT_ROUTING_MODE=smart` |
 | `fast` | DeepSeek | `deepseek-v4-flash` | `DEEPSEEK_FAST_MODEL` |
 | `complex` | DeepSeek | `deepseek-v4-pro` | `DEEPSEEK_PRO_MODEL` |
-| `manual` | DeepSeek/OpenAI/Qwen | user supplied | `AGENT_PROVIDER`, `LLM_MODEL` |
+| `manual` | DeepSeek/OpenAI/Qwen/Venice | user supplied | `AGENT_PROVIDER`, `LLM_MODEL` |
 
 Provider credentials:
 
@@ -282,6 +286,7 @@ Provider credentials:
 | OpenAI | `OPENAI_API_KEY` | `https://api.openai.com/v1` |
 | DeepSeek | `DEEPSEEK_API_KEY` | `https://api.deepseek.com/v1` |
 | Qwen | `QWEN_API_KEY` | `QWEN_BASE_URL` or DashScope compatible mode |
+| Venice | `VENICE_API_KEY` | `VENICE_API_BASE` or `https://api.venice.ai/api/v1` |
 
 Project-local credentials can be stored without committing secrets:
 

package/docs/auxiliary-image-generation.md

@@ -5,16 +5,18 @@ AgInTiFlow separates **skills** from **tools**:
 - A skill is instruction and routing context that teaches the agent when a capability is useful.
 - A tool is the deterministic callable function that performs the action and writes auditable artifacts.
 
-The optional image-generation skill is `image_generation`. Its tool is `generate_image`.
+The optional image-generation skills are `image_generation` and `venice_image_generation`. Both use the deterministic `generate_image` tool.
 
 ## Setup
 
-Store the GRS AI key project-locally:
+Store a GRS AI or Venice key project-locally:
 
 ```bash
 aginti login grsai
+aginti login venice
 # or
 printf '%s' "$GRSAI" | aginti keys set grsai --stdin
+printf '%s' "$VENICE_API_KEY" | aginti keys set venice --stdin
 ```
 
 Inside interactive chat, use either spelling:
@@ -24,7 +26,7 @@ Inside interactive chat, use either spelling:
 /auxiliary grsai
 ```
 
-The key is saved in `.aginti/.env` as `GRSAI` with `0600` permissions. The CLI and web app only report whether the key exists; they never return the raw value.
+Keys are saved in `.aginti/.env` as `GRSAI` or `VENICE_API_KEY` with `0600` permissions. The CLI and web app only report whether a key exists; they never return raw values.
 
 ## Runtime Flow
 
@@ -33,6 +35,8 @@ For image, cover, poster, illustration, photo, or logo-concept requests, the mod
 ```json
 {
   "prompt": "A polished cyan robot painting a circuit-board river, high-end product illustration",
+  "provider": "venice",
+  "model": "nano-banana-2",
   "outputDir": "artifacts/images/robot-cover",
   "outputStem": "robot-cover",
   "aspectRatio": "1:1",
@@ -40,18 +44,25 @@ For image, cover, poster, illustration, photo, or logo-concept requests, the mod
 }
 ```
 
-The tool uses the GRS AI Nano Banana API:
+With GRS AI, the tool uses the Nano Banana API:
 
 - `POST https://grsaiapi.com/v1/draw/nano-banana`
 - `POST https://grsaiapi.com/v1/draw/result`
 - `Authorization: Bearer <GRSAI>`
 
+With Venice, the tool uses the image-generation API:
+
+- `POST https://api.venice.ai/api/v1/image/generate`
+- `Authorization: Bearer <VENICE_API_KEY>`
+- image models such as `nano-banana-2`, `gpt-image-2`, `qwen-image-2`, `wan-2-7-text-to-image`, `bria-bg-remover`, and `venice-sd35`
+
 Saved workspace artifacts:
 
 - `prompt.txt`
 - `request_payload.redacted.json`
 - `submit_response.json`
 - `result_response.json`
+- `venice_result_response.json` for Venice image calls
 - `task_manifest.json`
 - generated image files, for example `image.png`
 

package/docs/model-selection.md

@@ -0,0 +1,70 @@
+# Model Selection
+
+AgInTiFlow treats model choice as a role-based control plane. A provider supplies models; a role decides how the agent uses one.
+
+## Default Roles
+
+| Role | CLI command | Default | Purpose |
+| --- | --- | --- | --- |
+| Route | `/route` | `deepseek/deepseek-v4-flash` | Fast planner, triage, short tasks, and routing decisions. |
+| Main | `/model` or `/main` | `deepseek/deepseek-v4-pro` | Complex executor for coding, debugging, writing, and long tasks. |
+| Spare | `/spare` | `openai/gpt-5.4` with `medium` reasoning | Optional fallback or cross-check model. |
+| Wrapper | `/wrapper` | `codex gpt-5.5 medium` | External coding assistant when wrapper tools are enabled. |
+| Auxiliary | `/auxilliary` | `grsai/nano-banana-2` | Image/media tools; Venice image models are optional. |
+
+Smart routing still works as before: normal work goes to the route model, and complex work goes to the main model. Manual provider/model selection remains available for one-off runs.
+
+## CLI Commands
+
+```bash
+aginti models
+aginti --list-models
+aginti --list-routes
+
+# one-shot overrides
+aginti --route-model deepseek-v4-flash --main-model deepseek-v4-pro "fix this project"
+aginti --provider venice --model venice-uncensored-1-2 --routing manual "draft a note"
+aginti --spare-provider openai --spare-model gpt-5.4 --spare-reasoning medium "review this design"
+aginti --allow-wrappers --wrapper codex --wrapper-model gpt-5.5 "patch this bug"
+aginti --image --aux-provider venice --aux-model gpt-image-2 "generate a logo"
+```
+
+Interactive commands:
+
+```text
+/models
+/route deepseek/deepseek-v4-flash
+/model deepseek/deepseek-v4-pro
+/spare openai/gpt-5.4 medium
+/wrapper codex gpt-5.5 medium
+/auxilliary model grsai/nano-banana-2
+```
+
+## Provider Buckets
+
+| Bucket | Provider | Typical use |
+| --- | --- | --- |
+| `deepseek` | DeepSeek | Default route/main because V4 Flash and V4 Pro are cheap and strong. |
+| `openai` | OpenAI | Spare/frontier checks, Codex-family work, and explicit manual routes. |
+| `qwen` | Qwen | Chinese and general-purpose OpenAI-compatible tasks. |
+| `venice-uncensored` | Venice | Venice-native uncensored text models. |
+| `venice-gpt` | Venice | GPT-family models through Venice. |
+| `venice-claude` | Venice | Claude-family models through Venice. |
+| `venice-gemma` | Venice | Gemma-family models through Venice. |
+| `venice-qwen` | Venice | Qwen-family models through Venice. |
+| `venice-image` | Venice | Image generation/editing such as Nano Banana, GPT Image, Wan, Qwen Image. |
+| `grsai` | GRS AI | Auxiliary image generation only. |
+
+## Keys
+
+Keys are set per project and stored only in ignored `.aginti/.env`:
+
+```bash
+aginti auth
+aginti auth venice
+aginti keys status
+printf '%s' "$VENICE_API_KEY" | aginti keys set venice --stdin
+```
+
+The web UI mirrors the same roles in its **Model roles** panel. Use the top provider/model fields for manual runs; use the role fields to change how smart routing chooses route, main, spare, wrapper, and auxiliary models.
+

package/package.json

@@ -1,8 +1,8 @@
 {
   "name": "@lazyingart/agintiflow",
-  "version": "0.17.6",
+  "version": "0.19.0",
   "type": "module",
-  "description": "AgInTiFlow is a resumable Playwright website-control agent with OpenAI-compatible tool calling.",
+  "description": "AgInTiFlow is a web-first coding agent and CLI with DeepSeek routing, sandboxed tools, model providers, canvas artifacts, and optional wrappers.",
   "license": "Apache-2.0",
   "homepage": "https://flow.lazying.art",
   "repository": {
@@ -18,6 +18,9 @@
     "browser",
     "playwright",
     "deepseek",
+    "venice",
+    "openai",
+    "qwen",
     "sandbox",
     "cli"
   ],
@@ -38,6 +41,8 @@
     "logos/banner-opaque.png",
     "logos/logo.png",
     "public/",
+    "references/*.md",
+    "references/run_design_scan.sh",
     "scripts/install-docker-ubuntu.sh",
     "scripts/setup-agent-toolchain-docker.sh",
     "scripts/real-deepseek-capabilities.js",
@@ -47,6 +52,7 @@
     "scripts/smoke-coding-tools.js",
     "scripts/smoke-capabilities.js",
     "scripts/smoke-inbox.js",
+    "scripts/smoke-model-roles.js",
     "scripts/smoke-platform.js",
     "scripts/smoke-skills.js",
     "scripts/smoke-tmux-tools.js",
@@ -73,11 +79,12 @@
     "smoke:skills": "node scripts/smoke-skills.js",
     "smoke:toolchain-docker": "node scripts/smoke-toolchain-docker.js",
     "smoke:inbox": "node scripts/smoke-inbox.js",
+    "smoke:model-roles": "node scripts/smoke-model-roles.js",
     "smoke:platform": "node scripts/smoke-platform.js",
     "smoke:tmux-tools": "node scripts/smoke-tmux-tools.js",
     "smoke:web-api": "node scripts/smoke-web-api.js",
     "real:deepseek": "node scripts/real-deepseek-capabilities.js",
-    "test": "npm run check && npm run smoke:web-api && npm run smoke:coding-tools && npm run smoke:auxiliary-tools && npm run smoke:auth && npm run smoke:capabilities && npm run smoke:platform && npm run smoke:skills && npm run smoke:tmux-tools && npm run smoke:cli-chat && npm run smoke:inbox",
+    "test": "npm run check && npm run smoke:web-api && npm run smoke:coding-tools && npm run smoke:auxiliary-tools && npm run smoke:auth && npm run smoke:capabilities && npm run smoke:model-roles && npm run smoke:platform && npm run smoke:skills && npm run smoke:tmux-tools && npm run smoke:cli-chat && npm run smoke:inbox",
     "pack:dry-run": "npm pack --dry-run",
     "smoke:capabilities": "node scripts/smoke-capabilities.js"
   },

package/public/app.js

@@ -22,9 +22,9 @@ const translations = {
     projectStatusTitle: "Project folder",
     setupTitle: "Provider setup",
     setupHelp:
-      "DeepSeek/OpenAI/Qwen keys are missing. Use mock mode, export an env var, or save a project-local model key.",
+      "DeepSeek/OpenAI/Qwen/Venice keys are missing. Use mock mode, export an env var, or save a project-local model key.",
     setupEnvHelp:
-      "Env vars: DEEPSEEK_API_KEY, OPENAI_API_KEY, QWEN_API_KEY, LLM_API_KEY, and optional GRSAI for image generation. Mock mode remains available for local tests.",
+      "Env vars: DEEPSEEK_API_KEY, OPENAI_API_KEY, QWEN_API_KEY, VENICE_API_KEY, LLM_API_KEY, and optional GRSAI for image generation. Mock mode remains available for local tests.",
     setupKeyLinksLabel: "Get keys:",
     deepseekKeyLink: "DeepSeek API keys",
     openaiKeyLink: "OpenAI API keys",
@@ -649,6 +649,21 @@ const languageField = document.querySelector("#language");
 const routingModeField = document.querySelector("#routingMode");
 const providerField = document.querySelector("#provider");
 const modelField = document.querySelector("#model");
+const modelOptionsEl = document.querySelector("#model-options");
+const modelCatalogEl = document.querySelector("#model-catalog");
+const modelRoleGridEl = document.querySelector("#model-role-grid");
+const modelRoutePillEl = document.querySelector("#model-route-pill");
+const routeProviderField = document.querySelector("#routeProvider");
+const routeModelField = document.querySelector("#routeModel");
+const mainProviderField = document.querySelector("#mainProvider");
+const mainModelField = document.querySelector("#mainModel");
+const spareProviderField = document.querySelector("#spareProvider");
+const spareModelField = document.querySelector("#spareModel");
+const spareReasoningField = document.querySelector("#spareReasoning");
+const wrapperModelField = document.querySelector("#wrapperModel");
+const wrapperReasoningField = document.querySelector("#wrapperReasoning");
+const auxiliaryProviderField = document.querySelector("#auxiliaryProvider");
+const auxiliaryModelField = document.querySelector("#auxiliaryModel");
 const routingHintEl = document.querySelector("#routing-hint");
 const modelRouteStatusEl = document.querySelector("#model-route-status");
 const projectStatusEl = document.querySelector("#project-status");
@@ -718,11 +733,16 @@ const defaults = {
   openai: "gpt-5.4-mini",
   deepseek: "deepseek-v4-flash",
   qwen: "qwen-plus",
+  venice: "venice-uncensored-1-2",
   mock: "mock-agent",
 };
 
 let currentLanguage = "en";
 let routingPresets = {};
+let modelCatalog = {};
+let modelRoles = {};
+let modelGroups = {};
+let auxiliaryModelCatalog = {};
 let taskProfiles = [];
 let projectInfo = null;
 let currentSessionId = "";
@@ -789,12 +809,14 @@ function renderKeyStatus(status = lastKeyStatus) {
     status.openai ? t("availableLabel") : t("missingLabel")
   } · DeepSeek ${status.deepseek ? t("availableLabel") : t("missingLabel")} · Qwen ${
     status.qwen ? t("availableLabel") : t("missingLabel")
+  } · Venice ${
+    status.venice ? t("availableLabel") : t("missingLabel")
   } · GRS AI ${
     status.grsai ? t("availableLabel") : t("missingLabel")
   } · ${t("mockLabel")} ${
     status.mock ? t("availableLabel") : t("missingLabel")
   }`;
-  if (setupCardEl) setupCardEl.hidden = Boolean(status.openai || status.deepseek || status.qwen);
+  if (setupCardEl) setupCardEl.hidden = Boolean(status.openai || status.deepseek || status.qwen || status.venice);
 }
 
 function renderProjectStatus(info = projectInfo) {
@@ -959,6 +981,90 @@ function renderSandboxLogs(logs) {
     .join("\n\n");
 }
 
+function providerModelOptions(provider = providerField.value) {
+  return modelCatalog[provider] || [];
+}
+
+function renderModelRoles() {
+  if (!modelRoleGridEl) return;
+  const roles = {
+    route: {
+      ...(modelRoles.route || {}),
+      provider: routeProviderField?.value || modelRoles.route?.provider || "deepseek",
+      model: routeModelField?.value || modelRoles.route?.model || "deepseek-v4-flash",
+    },
+    main: {
+      ...(modelRoles.main || {}),
+      provider: mainProviderField?.value || modelRoles.main?.provider || "deepseek",
+      model: mainModelField?.value || modelRoles.main?.model || "deepseek-v4-pro",
+    },
+    spare: {
+      ...(modelRoles.spare || {}),
+      provider: spareProviderField?.value || modelRoles.spare?.provider || "openai",
+      model: spareModelField?.value || modelRoles.spare?.model || "gpt-5.4",
+      reasoning: spareReasoningField?.value || modelRoles.spare?.reasoning || "medium",
+    },
+    wrapper: {
+      ...(modelRoles.wrapper || {}),
+      provider: preferredWrapperField?.value || modelRoles.wrapper?.provider || "codex",
+      model: wrapperModelField?.value || modelRoles.wrapper?.model || "gpt-5.5",
+      reasoning: wrapperReasoningField?.value || modelRoles.wrapper?.reasoning || "medium",
+    },
+    auxiliary: {
+      ...(modelRoles.auxiliary || {}),
+      provider: auxiliaryProviderField?.value || modelRoles.auxiliary?.provider || "grsai",
+      model: auxiliaryModelField?.value || modelRoles.auxiliary?.model || "nano-banana-2",
+    },
+  };
+  modelRoleGridEl.innerHTML = Object.values(roles)
+    .map((role) => {
+      const reasoning = role.reasoning ? ` · ${role.reasoning}` : "";
+      return `
+        <div class="model-role-card">
+          <strong>${escapeHtml(role.label || role.id)}</strong>
+          <span>${escapeHtml(`${role.provider}/${role.model}${reasoning}`)}</span>
+        </div>
+      `;
+    })
+    .join("");
+}
+
+function renderModelOptions() {
+  const provider = providerField.value || "deepseek";
+  const options = providerModelOptions(provider);
+  if (modelOptionsEl) {
+    modelOptionsEl.innerHTML = options
+      .map((item) => `<option value="${escapeHtml(item.id)}">${escapeHtml(item.label || item.id)}</option>`)
+      .join("");
+  }
+  if (modelRoutePillEl) {
+    const mode = routingModeField.value || "smart";
+    const primary = modelField.value.trim() || defaults[provider] || "";
+    const routeLabel = `${routeProviderField?.value || "deepseek"}/${routeModelField?.value || "deepseek-v4-flash"}`;
+    const mainLabel = `${mainProviderField?.value || "deepseek"}/${mainModelField?.value || "deepseek-v4-pro"}`;
+    modelRoutePillEl.textContent =
+      mode === "manual"
+        ? `${provider} · ${primary}`
+        : `${mode} · route ${routeLabel} · main ${mainLabel}`;
+  }
+  if (!modelCatalogEl) return;
+  if (options.length === 0) {
+    modelCatalogEl.innerHTML = "";
+    return;
+  }
+  modelCatalogEl.innerHTML = options
+    .map(
+      (item) => `
+        <button class="model-chip" type="button" data-model-id="${escapeHtml(item.id)}" data-active="${item.id === modelField.value.trim()}">
+          <strong>${escapeHtml(item.label || item.id)}</strong>
+          <span>${escapeHtml([item.bucket || item.role, item.context, item.reasoning?.length ? `reasoning ${item.reasoning.join("/")}` : ""].filter(Boolean).join(" · "))}</span>
+        </button>
+      `
+    )
+    .join("");
+  renderModelRoles();
+}
+
 function updateRoutingHint() {
   const mode = routingModeField.value || "smart";
   const hintKey = {
@@ -969,8 +1075,17 @@ function updateRoutingHint() {
   }[mode];
   routingHintEl.textContent = t(hintKey);
 
-  if (mode !== "manual") {
-    const preset = mode === "complex" ? routingPresets.complex : routingPresets.fast;
+  if (mode !== "manual" && providerField.value === "deepseek") {
+    const preset =
+      mode === "complex"
+        ? {
+            provider: mainProviderField?.value || routingPresets.complex?.provider,
+            model: mainModelField?.value || routingPresets.complex?.model,
+          }
+        : {
+            provider: routeProviderField?.value || routingPresets.fast?.provider,
+            model: routeModelField?.value || routingPresets.fast?.model,
+          };
     if (preset) {
       providerField.value = preset.provider === "deepseek" ? "deepseek" : providerField.value;
       modelField.value = preset.model || modelField.value;
@@ -980,6 +1095,7 @@ function updateRoutingHint() {
   modelRouteStatusEl.textContent = `${t("modelRouteStatus")}: ${providerField.value} / ${
     modelField.value.trim() || defaults[providerField.value] || ""
   }`;
+  renderModelOptions();
 }
 
 function updatePackageWarning() {
@@ -1037,6 +1153,17 @@ function formPayload() {
     routingMode: routingModeField.value,
     provider: providerField.value,
     model: modelField.value.trim(),
+    routeProvider: routeProviderField?.value || "deepseek",
+    routeModel: routeModelField?.value.trim() || "deepseek-v4-flash",
+    mainProvider: mainProviderField?.value || "deepseek",
+    mainModel: mainModelField?.value.trim() || "deepseek-v4-pro",
+    spareProvider: spareProviderField?.value || "openai",
+    spareModel: spareModelField?.value.trim() || "gpt-5.4",
+    spareReasoning: spareReasoningField?.value || "medium",
+    wrapperModel: wrapperModelField?.value.trim() || "gpt-5.5",
+    wrapperReasoning: wrapperReasoningField?.value || "medium",
+    auxiliaryProvider: auxiliaryProviderField?.value || "grsai",
+    auxiliaryModel: auxiliaryModelField?.value.trim() || "nano-banana-2",
     startUrl: document.querySelector("#startUrl").value.trim(),
     allowedDomains: document.querySelector("#allowedDomains").value.trim(),
     commandCwd: document.querySelector("#commandCwd").value.trim(),
@@ -2293,7 +2420,7 @@ deleteSessionButton.addEventListener("click", () => {
 });
 
 providerField.addEventListener("change", () => {
-  if (providerField.value === "mock") {
+  if (providerField.value === "mock" || providerField.value === "venice" || providerField.value === "openai" || providerField.value === "qwen") {
     routingModeField.value = "manual";
   }
 
@@ -2302,6 +2429,7 @@ providerField.addEventListener("change", () => {
     modelField.value === defaults.openai ||
     modelField.value === defaults.deepseek ||
     modelField.value === defaults.qwen ||
+    modelField.value === defaults.venice ||
     modelField.value === defaults.mock
   ) {
     modelField.value = defaults[providerField.value] || "";
@@ -2310,7 +2438,27 @@ providerField.addEventListener("change", () => {
   schedulePreferenceSave();
 });
 
+modelCatalogEl?.addEventListener("click", (event) => {
+  const button = event.target.closest("[data-model-id]");
+  if (!button) return;
+  modelField.value = button.dataset.modelId || modelField.value;
+  updateRoutingHint();
+  schedulePreferenceSave();
+});
+
 modelField.addEventListener("input", updateRoutingHint);
+[routeProviderField, routeModelField, mainProviderField, mainModelField, spareProviderField, spareModelField, spareReasoningField, wrapperModelField, wrapperReasoningField, auxiliaryProviderField, auxiliaryModelField]
+  .filter(Boolean)
+  .forEach((field) => {
+    field.addEventListener("input", () => {
+      renderModelRoles();
+      schedulePreferenceSave();
+    });
+    field.addEventListener("change", () => {
+      renderModelRoles();
+      schedulePreferenceSave();
+    });
+  });
 sandboxModeField.addEventListener("change", updatePackageWarning);
 packageInstallPolicyField.addEventListener("change", updatePackageWarning);
 allowWrapperToolsField.addEventListener("change", () => renderWrapperStatus());
@@ -2534,10 +2682,15 @@ async function loadConfig() {
   const data = await response.json();
   const prefs = data.preferences || {};
   routingPresets = data.routing?.presets || {};
+  modelCatalog = data.modelCatalog || {};
+  modelRoles = data.modelRoles || {};
+  modelGroups = data.modelGroups || {};
+  auxiliaryModelCatalog = data.auxiliaryModelCatalog || {};
   taskProfiles = data.taskProfiles || [];
   projectInfo = data.project || null;
   defaults.openai = data.defaults?.openai?.model || defaults.openai;
   defaults.qwen = data.defaults?.qwen?.model || defaults.qwen;
+  defaults.venice = data.defaults?.venice?.model || defaults.venice;
   defaults.deepseek = routingPresets.fast?.model || data.defaults?.deepseek?.model || defaults.deepseek;
   defaults.mock = data.defaults?.mock?.model || defaults.mock;
 
@@ -2546,6 +2699,17 @@ async function loadConfig() {
   routingModeField.value = prefs.routingMode || "smart";
   providerField.value = prefs.provider || "deepseek";
   modelField.value = prefs.model || defaults[providerField.value] || "deepseek-v4-flash";
+  if (routeProviderField) routeProviderField.value = prefs.routeProvider || modelRoles.route?.provider || "deepseek";
+  if (routeModelField) routeModelField.value = prefs.routeModel || modelRoles.route?.model || "deepseek-v4-flash";
+  if (mainProviderField) mainProviderField.value = prefs.mainProvider || modelRoles.main?.provider || "deepseek";
+  if (mainModelField) mainModelField.value = prefs.mainModel || modelRoles.main?.model || "deepseek-v4-pro";
+  if (spareProviderField) spareProviderField.value = prefs.spareProvider || modelRoles.spare?.provider || "openai";
+  if (spareModelField) spareModelField.value = prefs.spareModel || modelRoles.spare?.model || "gpt-5.4";
+  if (spareReasoningField) spareReasoningField.value = prefs.spareReasoning || modelRoles.spare?.reasoning || "medium";
+  if (wrapperModelField) wrapperModelField.value = prefs.wrapperModel || modelRoles.wrapper?.model || "gpt-5.5";
+  if (wrapperReasoningField) wrapperReasoningField.value = prefs.wrapperReasoning || modelRoles.wrapper?.reasoning || "medium";
+  if (auxiliaryProviderField) auxiliaryProviderField.value = prefs.auxiliaryProvider || modelRoles.auxiliary?.provider || "grsai";
+  if (auxiliaryModelField) auxiliaryModelField.value = prefs.auxiliaryModel || modelRoles.auxiliary?.model || "nano-banana-2";
   renderTaskProfiles(prefs.taskProfile || "auto");
   document.querySelector("#startUrl").value = prefs.startUrl || "";
   document.querySelector("#allowedDomains").value = prefs.allowedDomains || "";

package/public/index.html

@@ -45,16 +45,18 @@
           <div>
             <strong data-i18n="setupTitle">Provider setup</strong>
             <p class="subtle" data-i18n="setupHelp">
-              DeepSeek/OpenAI keys are missing. Use mock mode, export an env var, or save a project-local DeepSeek key.
+              DeepSeek/OpenAI/Qwen/Venice keys are missing. Use mock mode, export an env var, or save a project-local key.
             </p>
             <p class="subtle" data-i18n="setupEnvHelp">
-              Env vars: DEEPSEEK_API_KEY, OPENAI_API_KEY, QWEN_API_KEY, or LLM_API_KEY. Mock mode remains available for local tests.
+              Env vars: DEEPSEEK_API_KEY, OPENAI_API_KEY, QWEN_API_KEY, VENICE_API_KEY, or LLM_API_KEY. Mock mode remains available for local tests.
             </p>
             <p class="subtle key-links">
               <span data-i18n="setupKeyLinksLabel">Get keys:</span>
               <a href="https://platform.deepseek.com/api_keys" target="_blank" rel="noreferrer" data-i18n="deepseekKeyLink">DeepSeek API keys</a>
               <span aria-hidden="true">·</span>
               <a href="https://platform.openai.com/api-keys" target="_blank" rel="noreferrer" data-i18n="openaiKeyLink">OpenAI API keys</a>
+              <span aria-hidden="true">·</span>
+              <a href="https://venice.ai" target="_blank" rel="noreferrer">Venice API</a>
             </p>
           </div>
           <div class="grid">
@@ -64,6 +66,7 @@
                 <option value="deepseek">DeepSeek</option>
                 <option value="openai">OpenAI</option>
                 <option value="qwen">Qwen</option>
+                <option value="venice">Venice</option>
                 <option value="grsai">GRS AI image</option>
               </select>
             </label>
@@ -96,18 +99,92 @@
               <option value="deepseek">DeepSeek</option>
               <option value="openai">OpenAI</option>
               <option value="qwen">Qwen</option>
+              <option value="venice">Venice</option>
               <option value="mock" data-i18n="mockProviderOption">Mock local</option>
             </select>
             </label>
 
             <label>
               <span data-i18n="modelLabel">Model</span>
-              <input id="model" name="model" type="text" value="deepseek-v4-flash" />
+              <input id="model" name="model" type="text" list="model-options" value="deepseek-v4-flash" />
+              <datalist id="model-options"></datalist>
             </label>
           </div>
 
           <p id="routing-hint" class="subtle route-hint"></p>
           <p id="model-route-status" class="subtle route-hint"></p>
+          <section class="model-catalog-panel" aria-label="Model catalog">
+            <div class="model-catalog-head">
+              <span>Route</span>
+              <strong id="model-route-pill">DeepSeek flash/pro</strong>
+            </div>
+            <div id="model-role-grid" class="model-role-grid"></div>
+            <div id="model-catalog" class="model-catalog"></div>
+          </section>
+
+          <section class="model-role-editor" aria-label="Model role settings">
+            <div class="model-role-editor-head">
+              <strong>Model roles</strong>
+              <span>Route, main, spare, wrapper, auxiliary</span>
+            </div>
+            <div class="role-editor-grid">
+              <label>
+                <span>Route</span>
+                <select id="routeProvider" name="routeProvider">
+                  <option value="deepseek">DeepSeek</option>
+                  <option value="openai">OpenAI</option>
+                  <option value="qwen">Qwen</option>
+                  <option value="venice">Venice</option>
+                  <option value="mock">Mock</option>
+                </select>
+                <input id="routeModel" name="routeModel" type="text" placeholder="deepseek-v4-flash" />
+              </label>
+              <label>
+                <span>Main</span>
+                <select id="mainProvider" name="mainProvider">
+                  <option value="deepseek">DeepSeek</option>
+                  <option value="openai">OpenAI</option>
+                  <option value="qwen">Qwen</option>
+                  <option value="venice">Venice</option>
+                </select>
+                <input id="mainModel" name="mainModel" type="text" placeholder="deepseek-v4-pro" />
+              </label>
+              <label>
+                <span>Spare</span>
+                <select id="spareProvider" name="spareProvider">
+                  <option value="openai">OpenAI</option>
+                  <option value="deepseek">DeepSeek</option>
+                  <option value="qwen">Qwen</option>
+                  <option value="venice">Venice</option>
+                </select>
+                <input id="spareModel" name="spareModel" type="text" placeholder="gpt-5.4" />
+                <select id="spareReasoning" name="spareReasoning">
+                  <option value="low">low</option>
+                  <option value="medium">medium</option>
+                  <option value="high">high</option>
+                  <option value="xhigh">xhigh</option>
+                </select>
+              </label>
+              <label>
+                <span>Wrapper</span>
+                <input id="wrapperModel" name="wrapperModel" type="text" placeholder="gpt-5.5" />
+                <select id="wrapperReasoning" name="wrapperReasoning">
+                  <option value="low">low</option>
+                  <option value="medium">medium</option>
+                  <option value="high">high</option>
+                  <option value="xhigh">xhigh</option>
+                </select>
+              </label>
+              <label>
+                <span>Auxiliary</span>
+                <select id="auxiliaryProvider" name="auxiliaryProvider">
+                  <option value="grsai">GRS AI</option>
+                  <option value="venice">Venice image</option>
+                </select>
+                <input id="auxiliaryModel" name="auxiliaryModel" type="text" placeholder="nano-banana-2" />
+              </label>
+            </div>
+          </section>
 
           <label>
             <span data-i18n="taskProfileLabel">Task profile</span>