kivi-cli 2.0.6__tar.gz → 2.0.7__tar.gz

{kivi_cli-2.0.6/src/kivi_cli.egg-info → kivi_cli-2.0.7}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: kivi-cli
-Version: 2.0.6
+Version: 2.0.7
 Summary: Kivi AI v2 Backend — Unified FastAPI gateway for AI coding agents
 Author: santhosh_bytes
 License: Proprietary

{kivi_cli-2.0.6 → kivi_cli-2.0.7}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "kivi-cli"
-version = "2.0.6"
+version = "2.0.7"
 description = "Kivi AI v2 Backend — Unified FastAPI gateway for AI coding agents"
 requires-python = ">=3.10"
 license = {text = "Proprietary"}

{kivi_cli-2.0.6 → kivi_cli-2.0.7/src/kivi_cli.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: kivi-cli
-Version: 2.0.6
+Version: 2.0.7
 Summary: Kivi AI v2 Backend — Unified FastAPI gateway for AI coding agents
 Author: santhosh_bytes
 License: Proprietary

{kivi_cli-2.0.6 → kivi_cli-2.0.7}/src/kivi_cli.egg-info/SOURCES.txt

@@ -17,6 +17,7 @@ src/kiviai/services/providers/__init__.py
 src/kiviai/services/providers/base.py
 src/kiviai/services/providers/claude.py
 src/kiviai/services/providers/copilot.py
+src/kiviai/services/providers/gemini.py
 src/kiviai/services/providers/inhouse.py
 src/kiviai/services/providers/kivi.py
 src/kiviai/services/providers/model_costs.py

{kivi_cli-2.0.6 → kivi_cli-2.0.7}/src/kiviai/backend.py

@@ -33,7 +33,7 @@ from kiviai.services.git_service import GitService
 logging.basicConfig(level=logging.INFO, format="%(asctime)s [%(levelname)s] %(message)s")
 log = logging.getLogger("kiviai")
 
-app = FastAPI(title="Kivi AI v2", version="2.0.6")
+app = FastAPI(title="Kivi AI v2", version="2.0.7")
 app.add_middleware(CORSMiddleware, allow_origins=["*"], allow_methods=["*"], allow_headers=["*"])
 
 HERE = os.path.dirname(os.path.abspath(__file__))
@@ -161,7 +161,7 @@ async def _run_task(task: dict):
         "opencode":    "opencode",
         "copilot":     "copilot",
         "local":       "vllm",
-
+        "gemini":      "gemini",
     }
 
     # Map short model names to current Claude model IDs.
@@ -189,6 +189,13 @@ async def _run_task(task: dict):
         if "vllm_model" not in extra:
             extra["vllm_model"] = ""
 
+    # For "gemini" agent, pass gemini model from extra
+    if model_name == "gemini":
+        extra = task.setdefault("extra", {})
+        if extra.get("gemini_model"):
+            extra["model_override"] = extra["gemini_model"]
+        if extra.get("gemini_api_key"):
+            extra["api_key"] = extra["gemini_api_key"]
 
     log.info("_run_task starting: agent=%s provider=%s prompt=%s", model_name, provider_name, task["prompt"][:60])
     try:
@@ -1254,6 +1261,7 @@ _AGENTS = [
     {"id": "opencode",    "name": "OpenCode",    "provider": "opencode", "description": "Open-source agentic coding CLI"},
     {"id": "copilot",     "name": "Copilot",     "provider": "copilot", "description": "GitHub Copilot SDK agent"},
     {"id": "local",       "name": "Local",        "provider": "vllm", "description": "Local vLLM model (Qwen 3.5-9B)"},
+    {"id": "gemini",      "name": "Gemini",       "provider": "gemini", "description": "Google Gemini — fast streaming, great for brainstorming"},
 ]
 
 

{kivi_cli-2.0.6 → kivi_cli-2.0.7}/src/kiviai/index.html

@@ -3190,6 +3190,7 @@ textarea:focus-visible {
                 <option value="claude-code">Claude Code</option>
                 <option value="opencode">OpenCode</option>
                 <option value="copilot">Copilot</option>
+                <option value="gemini">Gemini</option>
               </select>
               <select id="ag-model"></select>
               <input class="ag-cwd" id="ag-cwd" placeholder="~/project" autocomplete="off" />
@@ -3227,6 +3228,7 @@ textarea:focus-visible {
             <option value="claude-code">Claude Code</option>
             <option value="opencode">OpenCode</option>
             <option value="copilot">Copilot</option>
+            <option value="gemini">Gemini</option>
           </select>
           <div class="ctrl-sep"></div>
           <div class="ctrl-group">
@@ -3494,20 +3496,40 @@ textarea:focus-visible {
 </table>
 </div>
 
+<h3 class="docs-h3">Gemini CLI flags — supported via UI</h3>
+<div class="docs-table-wrap">
+<table class="docs-table">
+<thead><tr><th>UI Control</th><th>Equivalent CLI flag</th><th>Notes</th></tr></thead>
+<tbody>
+<tr><td>Agent dropdown → Gemini (planned)</td><td><code>gemini -p "…"</code></td><td>Provider not wired yet — see roadmap</td></tr>
+<tr><td>Model picker</td><td><code>-m gemini-2.5-pro</code></td><td>Any Gemini model string</td></tr>
+<tr><td>Mode → plan</td><td><code>--approval-mode plan</code></td><td>Read-only, no writes</td></tr>
+<tr><td>Mode → auto</td><td><code>--approval-mode auto_edit</code></td><td>Auto-approve edits, prompt for rest</td></tr>
+<tr><td>Mode → bypass</td><td><code>--approval-mode yolo</code></td><td>Auto-approve everything</td></tr>
+<tr><td>Resume session</td><td><code>-r latest</code></td><td>Gemini uses index-based resume</td></tr>
+<tr><td>CWD / extra dirs</td><td><code>--include-directories</code></td><td>Comma-separated extra workspace dirs</td></tr>
+<tr><td>Sandbox toggle (planned)</td><td><code>-s --sandbox</code></td><td>Isolated execution environment</td></tr>
+</tbody>
+</table>
+</div>
 
 <h3 class="docs-h3">Feature Roadmap — CLI parity gaps to close</h3>
 <div class="docs-table-wrap">
 <table class="docs-table">
 <thead><tr><th>Feature</th><th>CLI Source</th><th>Priority</th><th>Status</th></tr></thead>
 <tbody>
+<tr><td>Gemini provider wiring</td><td>Gemini CLI <code>-p</code> headless</td><td>High</td><td class="status-done">Done</td></tr>
+<tr><td>Session fork UI</td><td>OpenCode <code>--fork</code>, Claude <code>--fork-session</code></td><td>High</td><td class="status-done">Done</td></tr>
 <tr><td>Thinking blocks toggle</td><td>OpenCode <code>--thinking</code></td><td>Medium</td><td class="status-done">Done</td></tr>
 <tr><td>Effort level selector</td><td>Claude <code>--effort low/medium/high/xhigh/max</code></td><td>Medium</td><td class="status-done">Done</td></tr>
 <tr><td>Max budget USD input</td><td>Claude <code>--max-budget-usd</code></td><td>Medium</td><td class="status-done">Done</td></tr>
 <tr><td>Allowed/disallowed tools</td><td>Claude <code>--allowedTools / --disallowedTools</code></td><td>Medium</td><td class="status-done">Done</td></tr>
 <tr><td>Custom system prompt</td><td>Claude <code>--system-prompt / --append-system-prompt</code></td><td>High</td><td class="status-done">Done</td></tr>
-<tr><td>MCP server config panel</td><td>Claude <code>--mcp-config</code></td><td>High</td><td class="status-todo">TODO</td></tr>
+<tr><td>MCP server config panel</td><td>Claude/Gemini <code>--mcp-config</code></td><td>High</td><td class="status-todo">TODO</td></tr>
 <tr><td>File attachments on prompt</td><td>OpenCode <code>-f --file</code>, Claude <code>--file</code></td><td>High</td><td class="status-done">Done</td></tr>
-<tr><td>Worktree per task</td><td>Claude <code>-w --worktree</code></td><td>Medium</td><td class="status-todo">TODO</td></tr>
+<tr><td>Worktree per task</td><td>Claude/Gemini <code>-w --worktree</code></td><td>Medium</td><td class="status-todo">TODO</td></tr>
+<tr><td>Sandbox toggle</td><td>Gemini <code>-s --sandbox</code></td><td>Low</td><td class="status-todo">TODO</td></tr>
+<tr><td>Session export/import</td><td>OpenCode <code>export / import</code></td><td>Medium</td><td class="status-todo">TODO</td></tr>
 <tr><td>OpenCode stats panel</td><td>OpenCode <code>stats</code> (token usage/cost history)</td><td>Low</td><td class="status-todo">TODO</td></tr>
 <tr><td>PR auto-run (opencode pr)</td><td>OpenCode <code>pr &lt;number&gt;</code></td><td>Medium</td><td class="status-todo">TODO</td></tr>
 <tr><td>Output format (JSON/stream)</td><td>Claude <code>--output-format stream-json</code></td><td>Low</td><td class="status-done">Done (internal)</td></tr>
@@ -5080,6 +5102,7 @@ const AGENT_MODELS = {
     { v: 'anthropic/claude-sonnet-5',      label: 'anthropic/claude-sonnet-5' },
     { v: 'anthropic/claude-opus-4-8',      label: 'anthropic/claude-opus-4-8' },
     { v: 'openai/gpt-5.4',                label: 'openai/gpt-5.4' },
+    { v: 'google/gemini-2.5-pro',          label: 'google/gemini-2.5-pro' },
   ],
   'copilot': [
     { v: 'gpt-5.4',           label: 'gpt-5.4' },
@@ -5094,6 +5117,14 @@ const AGENT_MODELS = {
     { v: 'o3-mini',           label: 'o3-mini' },
     { v: 'o4-mini',           label: 'o4-mini' },
   ],
+  'gemini': [
+    { v: 'gemini-2.5-pro',       label: 'gemini-2.5-pro' },
+    { v: 'gemini-2.5-flash',     label: 'gemini-2.5-flash' },
+    { v: 'gemini-2.0-flash',     label: 'gemini-2.0-flash' },
+    { v: 'gemini-2.0-flash-lite', label: 'gemini-2.0-flash-lite' },
+    { v: 'gemini-1.5-pro',       label: 'gemini-1.5-pro' },
+    { v: 'gemini-1.5-flash',     label: 'gemini-1.5-flash' },
+  ],
 };
 
 function renderModelOptions(models) {
@@ -5210,6 +5241,10 @@ function buildLaunchBody({ prompt, agent, selectedModel, cwd, extras }) {
     body.extra.vllm_model = selectedModel || '';
     body.extra.vllm_key = localStorage.getItem('vllm_key') || 'dummy';
   }
+  if (agent === 'gemini') {
+    const gKey = localStorage.getItem('gemini_api_key');
+    if (gKey) body.extra.api_key = gKey;
+    if (selectedModel) body.extra.gemini_model = selectedModel;
   }
   return body;
 }

{kivi_cli-2.0.6 → kivi_cli-2.0.7}/src/kiviai/services/providers/base.py

@@ -2,7 +2,7 @@
 Base provider protocol for all AI integrations.
 
 Every provider implements this interface so the system can swap
-between Claude, Copilot, VLLM, etc. with zero friction.
+between Claude, Gemini, Copilot, VLLM, etc. with zero friction.
 """
 from __future__ import annotations
 

kivi_cli-2.0.7/src/kiviai/services/providers/gemini.py

@@ -0,0 +1,120 @@
+"""Google Gemini provider — uses google-genai SDK."""
+from __future__ import annotations
+
+import asyncio
+import json
+import os
+from typing import AsyncIterator
+
+from .base import BaseProvider, ProviderConfig, ProviderEvent, EventType
+
+
+class GeminiProvider(BaseProvider):
+    name = "gemini"
+    display_name = "Google Gemini"
+    description = "Google's multimodal AI. Fast streaming text generation via the Gemini API. Best for quick questions, summaries, and brainstorming. Requires GEMINI_API_KEY."
+    supports_streaming = True
+    supports_tools = False
+    supports_sessions = False
+    available_models = [
+        "gemini-2.5-pro", "gemini-2.5-flash", "gemini-2.0-flash",
+        "gemini-2.0-flash-lite", "gemini-1.5-pro", "gemini-1.5-flash",
+    ]
+
+    def __init__(self):
+        self._stop = False
+
+    async def run(self, prompt: str, config: ProviderConfig, history=None) -> AsyncIterator[ProviderEvent]:
+        self._stop = False
+
+        api_key = config.api_key or os.environ.get("GEMINI_API_KEY") or os.environ.get("GOOGLE_API_KEY", "")
+        model = config.model or "gemini-2.5-flash"
+
+        if not api_key:
+            yield ProviderEvent(type=EventType.ERROR, content="GEMINI_API_KEY not set")
+            return
+
+        try:
+            from google import genai
+            from google.genai import types
+
+            client = genai.Client(api_key=api_key)
+
+            contents = []
+            if history:
+                for h in history:
+                    if h.get("role") in ("user", "assistant") and not h.get("metadata"):
+                        role = "user" if h["role"] == "user" else "model"
+                        contents.append(types.Content(role=role, parts=[types.Part(text=h["content"])]))
+            contents.append(types.Content(role="user", parts=[types.Part(text=prompt)]))
+
+            response = await asyncio.to_thread(
+                lambda: client.models.generate_content_stream(
+                    model=model,
+                    contents=contents,
+                    config=types.GenerateContentConfig(
+                        temperature=0.7,
+                        max_output_tokens=4096,
+                    ),
+                )
+            )
+
+            text_buf = ""
+            total_input_tokens = 0
+            total_output_tokens = 0
+            for chunk in response:
+                if self._stop:
+                    yield ProviderEvent(type=EventType.TEXT, content="⏹ *stopped*")
+                    yield ProviderEvent(type=EventType.DONE)
+                    return
+
+                if chunk.text:
+                    text_buf += chunk.text
+                    yield ProviderEvent(type=EventType.TEXT, content=chunk.text)
+
+                # Capture usage metadata (available in the final chunk or via usage_metadata)
+                if hasattr(chunk, 'usage_metadata') and chunk.usage_metadata:
+                    um = chunk.usage_metadata
+                    total_input_tokens += getattr(um, 'prompt_token_count', 0) or 0
+                    total_output_tokens += getattr(um, 'candidates_token_count', 0) or 0
+
+        except ImportError:
+            yield ProviderEvent(
+                type=EventType.ERROR,
+                content="google-genai package required: pip install google-genai",
+            )
+            return
+        except Exception as e:
+            yield ProviderEvent(type=EventType.ERROR, content=str(e))
+            return
+
+        # Emit cost event with calculated cost from usage
+        from .model_costs import estimate_cost
+        usage = {
+            "input_tokens": total_input_tokens,
+            "output_tokens": total_output_tokens,
+        }
+        total_cost = estimate_cost(
+            model,
+            input_tokens=total_input_tokens,
+            output_tokens=total_output_tokens,
+        )
+        yield ProviderEvent(
+            type=EventType.COST,
+            metadata={
+                "total_cost_usd": total_cost,
+                "usage": usage,
+            },
+        )
+
+        yield ProviderEvent(type=EventType.DONE)
+
+    async def stop(self):
+        self._stop = True
+
+    async def health_check(self):
+        api_key = os.environ.get("GEMINI_API_KEY") or os.environ.get("GOOGLE_API_KEY", "")
+        return {
+            "status": "ok" if api_key else "no_api_key",
+            "provider": self.name,
+        }

{kivi_cli-2.0.6 → kivi_cli-2.0.7}/src/kiviai/services/providers/model_costs.py

@@ -51,6 +51,17 @@ COSTS: dict[str, tuple[float, float]] = {
     "o3-mini":                         (1.10,  4.40),  # O3 Mini
     "o4-mini":                         (1.10,  4.40),  # O4 Mini
 
+    # ── Google Gemini ───────────────────────────────────────────
+    # Model                                         (input, output)
+    "gemini-3.1-pro-preview":          (1.25, 10.00),  # Gemini 3.1 Pro
+    "gemini-2.5-pro":                  (1.25, 10.00),  # Gemini 2.5 Pro
+    "gemini-2.5-flash":                (0.30,  2.50),  # Gemini 2.5 Flash
+    "gemini-2.5-flash-lite":           (0.08,  0.60),  # Gemini 2.5 Flash Lite (free tier eligible)
+    "gemini-2.0-flash":                (0.10,  1.00),  # Gemini 2.0 Flash
+    "gemini-2.0-flash-lite":           (0.08,  0.60),  # Gemini 2.0 Flash Lite
+    "gemini-1.5-pro":                  (1.25,  5.00),  # Gemini 1.5 Pro
+    "gemini-1.5-flash":                (0.08,  0.30),  # Gemini 1.5 Flash
+
     # ── OpenCode model aliases (full paths) ─────────────────────
     "anthropic/claude-sonnet-4-20250514":     (3.00, 15.00),
     "anthropic/claude-opus-4-20250514":       (15.00, 75.00),
@@ -59,6 +70,8 @@ COSTS: dict[str, tuple[float, float]] = {
     "openai/gpt-4.1-mini":                   (0.40,  1.60),
     "openai/o3":                             (10.50, 42.00),
     "openai/o4-mini":                        (1.10,  4.40),
+    "google/gemini-2.5-flash":               (0.30,  2.50),
+    "google/gemini-2.5-pro":                 (1.25, 10.00),
 }
 
 
@@ -83,7 +96,7 @@ def estimate_cost(
         output_tokens: Generated/completion tokens produced
         cache_read_tokens: Tokens served from cache (discounted)
         cache_write_tokens: Tokens written to cache (premium)
-        reasoning_tokens: Tokens spent on internal reasoning (GPT-o series)
+        reasoning_tokens: Tokens spent on internal reasoning (GPT-o, Gemini)
 
     Returns:
         Estimated cost in USD (0.0 for unknown/local models)

{kivi_cli-2.0.6 → kivi_cli-2.0.7}/src/kiviai/services/providers/opencode.py

@@ -31,7 +31,7 @@ class OpenCodeProvider(BaseProvider):
         # Below require the user's own provider credentials (`opencode auth login`).
         "anthropic/claude-sonnet-5", "anthropic/claude-opus-4-8", "anthropic/claude-haiku-4-5",
         "openai/gpt-5.4", "openai/gpt-4.1", "openai/gpt-4.1-mini", "openai/o3", "openai/o4-mini",
-        
+        "google/gemini-2.5-flash", "google/gemini-2.5-pro",
     ]
 
     def __init__(self):

{kivi_cli-2.0.6 → kivi_cli-2.0.7}/src/kiviai/services/providers/registry.py

@@ -85,5 +85,11 @@ def _auto_register():
     except ImportError as e:
         log.debug("vLLM provider not available: %s", e)
 
+    try:
+        from .gemini import GeminiProvider
+        register("gemini", GeminiProvider, aliases=["google", "gemini-pro", "gemini-flash"])
+    except ImportError as e:
+        log.debug("Gemini provider not available: %s", e)
+
 
 _auto_register()