infernoflow 0.10.7 → 0.10.8

package/README.md

@@ -166,6 +166,7 @@ infernoflow doc-gate --json
 | `infernoflow status` | At-a-glance health of your contract |
 | `infernoflow suggest` | Generate an AI prompt, apply capability updates |
 | `infernoflow implement` | Generate implementation prompts for coding agents |
+| `infernoflow run` | One-command local-model flow with validation and rollback |
 | `infernoflow pr-impact` | Analyze changed files and infer capability/doc drift |
 | `infernoflow sync --auto` | Deterministic sync flow for agents (skeleton) |
 | `infernoflow check` | Full validation: contract, capabilities, scenarios, changelog |
@@ -185,6 +186,9 @@ infernoflow implement "..." --mode both
 infernoflow implement "..." --mode cursor
 infernoflow implement "..." --mode generic
 infernoflow implement "..." --mode both --copy
+infernoflow run "add favorite badge to tasks"
+infernoflow run "sync check" --dry-run
+infernoflow run "sync check" --json
 infernoflow pr-impact
 infernoflow pr-impact --json
 infernoflow sync --auto
@@ -232,6 +236,43 @@ Proposed Changes
 
 Works with any AI — Claude, ChatGPT, GitHub Copilot, Cursor, or your own setup.
 
+## `infernoflow run` — zero copy/paste flow
+
+Run one command and infernoflow will:
+1. Detect drift (`pr-impact`)
+2. Generate suggestion via local model
+3. Apply inferno updates
+4. Validate with `check`
+5. Roll back automatically if validation fails
+
+```bash
+infernoflow run "add favorite badge to tasks and filter by favorite"
+```
+
+Machine mode:
+
+```bash
+infernoflow run "sync check" --json
+```
+
+Local model configuration (required):
+
+```bash
+# default provider: ollama
+set INFERNO_LOCAL_PROVIDER=ollama
+set INFERNO_LOCAL_ENDPOINT=http://127.0.0.1:11434/api/generate
+set INFERNO_LOCAL_MODEL=llama3.1:8b
+```
+
+Optional OpenAI-compatible local server:
+
+```bash
+set INFERNO_LOCAL_PROVIDER=openai
+set INFERNO_LOCAL_ENDPOINT=http://127.0.0.1:1234/v1/chat/completions
+set INFERNO_LOCAL_MODEL=local-model
+set INFERNO_LOCAL_API_KEY=local
+```
+
 ## `infernoflow implement` — code-agent execution prompts
 
 Generate coding prompts from your project context and inferno contract:
@@ -279,11 +320,12 @@ Recommended chain:
 
 ```yaml
 # .github/workflows/ci.yml
-- name: infernoflow impact + check
-  run: |
-    npx infernoflow pr-impact --json
-    npx infernoflow check --json
+- name: infernoflow run
+  run: npx infernoflow run "sync check" --json
   env:
+    INFERNO_LOCAL_PROVIDER: ollama
+    INFERNO_LOCAL_ENDPOINT: http://127.0.0.1:11434/api/generate
+    INFERNO_LOCAL_MODEL: llama3.1:8b
     BASE_SHA: ${{ github.event.pull_request.base.sha }}
     HEAD_SHA: ${{ github.event.pull_request.head.sha }}
 ```

package/bin/infernoflow.mjs

@@ -13,6 +13,7 @@ const COMMAND_DESCRIPTIONS = {
   status: "Show contract health at a glance",
   "pr-impact": "Summarize PR impact on capabilities and docs",
   sync: "Run deterministic inferno sync flow",
+  run: "One-command detect/propose/apply/validate flow",
   "doc-gate": "Fail if code changed but docs were not updated",
   suggest: "Generate AI prompt + apply capability updates",
   implement: "Generate code-agent implementation prompt(s)",
@@ -25,6 +26,7 @@ const COMMAND_HANDLERS = {
   status: async (args) => (await import("../lib/commands/status.mjs")).statusCommand(args),
   "pr-impact": async (args) => (await import("../lib/commands/prImpact.mjs")).prImpactCommand(args),
   sync: async (args) => (await import("../lib/commands/syncAuto.mjs")).syncCommand(args),
+  run: async (args) => (await import("../lib/commands/run.mjs")).runCommand(args),
   suggest: async (args) => (await import("../lib/commands/suggest.mjs")).suggestCommand(args),
   implement: async (args) => (await import("../lib/commands/implement.mjs")).implementCommand(args),
   context: async (args) => (await import("../lib/commands/context.mjs")).contextCommand(args),
@@ -70,6 +72,11 @@ ${formatCommandsHelp()}
     --mode <type>       cursor | generic | both (default: both)
     --copy, -c          Copy generated prompt(s) to clipboard
 
+  ${bold("run options:")}
+    --dry-run           Execute full flow without writing files
+    --json              Emit machine-readable events and result payload
+    --no-rollback       Keep changes even if validation fails
+
   ${bold("Typical workflow:")}
     ${gray('1. infernoflow context --intent "what I want to build"')}
     ${gray("2. [paste inferno/CONTEXT.md into Claude / Cursor / Copilot]")}
@@ -83,6 +90,7 @@ ${formatCommandsHelp()}
     ${gray("doc-gate --json")}
     ${gray("pr-impact --json")}
     ${gray("sync --auto --json")}
+    ${gray('run "task" --json')}
 `;
 
 const [, , cmd, ...rest] = process.argv;

package/lib/ai/localProvider.mjs

@@ -0,0 +1,88 @@
+const DEFAULT_TIMEOUT_MS = 45000;
+
+function withTimeout(ms) {
+  const controller = new AbortController();
+  const timer = setTimeout(() => controller.abort(), ms);
+  return { controller, timer };
+}
+
+async function callOllama(prompt, timeoutMs) {
+  const endpoint = process.env.INFERNO_LOCAL_ENDPOINT || "http://127.0.0.1:11434/api/generate";
+  const model = process.env.INFERNO_LOCAL_MODEL || "llama3.1:8b";
+  const { controller, timer } = withTimeout(timeoutMs);
+  try {
+    const res = await fetch(endpoint, {
+      method: "POST",
+      headers: { "Content-Type": "application/json" },
+      signal: controller.signal,
+      body: JSON.stringify({
+        model,
+        prompt,
+        stream: false,
+      }),
+    });
+    if (!res.ok) {
+      const body = await res.text();
+      throw new Error(`local_model_http_${res.status}: ${body.slice(0, 240)}`);
+    }
+    const data = await res.json();
+    if (!data?.response || typeof data.response !== "string") {
+      throw new Error("local_model_invalid_response");
+    }
+    return data.response.trim();
+  } finally {
+    clearTimeout(timer);
+  }
+}
+
+async function callOpenAICompat(prompt, timeoutMs) {
+  const endpoint = process.env.INFERNO_LOCAL_ENDPOINT || "http://127.0.0.1:1234/v1/chat/completions";
+  const model = process.env.INFERNO_LOCAL_MODEL || "local-model";
+  const apiKey = process.env.INFERNO_LOCAL_API_KEY || "local";
+  const { controller, timer } = withTimeout(timeoutMs);
+  try {
+    const res = await fetch(endpoint, {
+      method: "POST",
+      headers: {
+        "Content-Type": "application/json",
+        Authorization: `Bearer ${apiKey}`,
+      },
+      signal: controller.signal,
+      body: JSON.stringify({
+        model,
+        temperature: 0.1,
+        messages: [
+          { role: "system", content: "Return JSON only." },
+          { role: "user", content: prompt },
+        ],
+      }),
+    });
+    if (!res.ok) {
+      const body = await res.text();
+      throw new Error(`local_model_http_${res.status}: ${body.slice(0, 240)}`);
+    }
+    const data = await res.json();
+    const text = data?.choices?.[0]?.message?.content;
+    if (!text || typeof text !== "string") {
+      throw new Error("local_model_invalid_response");
+    }
+    return text.trim();
+  } finally {
+    clearTimeout(timer);
+  }
+}
+
+export async function generateWithLocalModel(prompt, options = {}) {
+  if (process.env.INFERNO_LOCAL_MOCK_RESPONSE) {
+    return process.env.INFERNO_LOCAL_MOCK_RESPONSE;
+  }
+
+  const provider = (process.env.INFERNO_LOCAL_PROVIDER || "ollama").toLowerCase();
+  const timeoutMs = Number(options.timeoutMs || process.env.INFERNO_LOCAL_TIMEOUT_MS || DEFAULT_TIMEOUT_MS);
+
+  if (provider === "openai") {
+    return callOpenAICompat(prompt, timeoutMs);
+  }
+  return callOllama(prompt, timeoutMs);
+}
+