Pārlūkot izejas kodu

evals: add Gemini 2.5 Flash backend

Drew Ritter 4 mēneši atpakaļ
vecāks
revīzija
9efbb7dd0d

+ 2 - 1
evals/README.md

@@ -64,7 +64,8 @@ uv run drill list
 | `claude-opus-4-6-1m` | Claude Code | opus-4-6 (1M context) |
 | `claude-opus-4-7-1m` | Claude Code | opus-4-7 (1M context) |
 | `codex` | Codex CLI | — |
-| `gemini` | Gemini CLI | — |
+| `gemini` | Gemini CLI | auto-gemini-3 |
+| `gemini-2-5-flash` | Gemini CLI | gemini-2.5-flash |
 
 ## Project structure
 

+ 23 - 0
evals/backends/gemini-2-5-flash.yaml

@@ -0,0 +1,23 @@
+name: gemini-2-5-flash
+cli: gemini
+args:
+  - "--yolo"
+  - "-m"
+  - "gemini-2.5-flash"
+required_env: []
+hooks:
+  pre_run:
+    - link_gemini_extension
+  post_run: []
+shutdown: "/exit"
+idle:
+  quiescence_seconds: 5
+  ready_pattern: "Type your message|^\\s*>"
+busy_pattern: "Thinking\\.\\.\\.|Executing"
+startup_timeout: 60
+turn_timeout: 300
+terminal:
+  cols: 200
+  rows: 50
+session_logs:
+  pattern: "~/.gemini/tmp/*/chats/session-*.json"

+ 1 - 1
evals/backends/gemini.yaml

@@ -3,7 +3,7 @@ cli: gemini
 args:
   - "--yolo"
   - "-m"
-  - "gemini-2.5-flash"
+  - "auto-gemini-3"
 required_env: []
 hooks:
   pre_run:

+ 11 - 0
evals/tests/test_backend.py

@@ -33,6 +33,17 @@ class TestLoadBackend:
         assert backend.family == "claude"
         assert backend.model == "claude-opus-4-6"
 
+    def test_loads_gemini_default_and_flash_variant(self, backends_dir):
+        backend = load_backend("gemini", backends_dir)
+        assert backend.name == "gemini"
+        assert backend.family == "gemini"
+        assert backend.model == "auto-gemini-3"
+
+        flash_backend = load_backend("gemini-2-5-flash", backends_dir)
+        assert flash_backend.name == "gemini-2-5-flash"
+        assert flash_backend.family == "gemini"
+        assert flash_backend.model == "gemini-2.5-flash"
+
 
 class TestBackendBuildCommand:
     def test_claude_build_command(self, backends_dir, monkeypatch):