voice-vibecoder 2.24.1__tar.gz → 2.26.0__tar.gz

{voice_vibecoder-2.24.1 → voice_vibecoder-2.26.0}/CHANGELOG.md

@@ -2,6 +2,28 @@
 
 <!-- version list -->
 
+## v2.26.0 (2026-07-13)
+
+### Features
+
+- **providers**: Pass the configured language to transcription
+  ([`cc06dcb`](https://github.com/snokam/voice-vibecoder/commit/cc06dcb45e9844827cfa2792b88ee3561f8c6a38))
+
+- **tools**: Merge a branch into main by voice
+  ([`8fbddd6`](https://github.com/snokam/voice-vibecoder/commit/8fbddd6da288634e4ab3a80da9d86cba9c1d1e8f))
+
+
+## v2.25.0 (2026-07-13)
+
+### Features
+
+- **session**: Resume conversations with recent history in context
+  ([`a4a60b2`](https://github.com/snokam/voice-vibecoder/commit/a4a60b214c68e028111d873fc6c1d33f47d93967))
+
+- **voice**: Follow the user's language instead of locking to one
+  ([`b828007`](https://github.com/snokam/voice-vibecoder/commit/b82800721340f96aaaa29ada12276c0466eeca2a))
+
+
 ## v2.24.1 (2026-07-13)
 
 ### Bug Fixes

{voice_vibecoder-2.24.1 → voice_vibecoder-2.26.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: voice-vibecoder
-Version: 2.24.1
+Version: 2.26.0
 Summary: Olaf The Vibecoder — a voice-controlled coding assistant using OpenAI Realtime API + Claude Code
 Project-URL: Homepage, https://github.com/snokam/voice-vibecoder
 Project-URL: Repository, https://github.com/snokam/voice-vibecoder

{voice_vibecoder-2.24.1 → voice_vibecoder-2.26.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "voice-vibecoder"
-version = "2.24.1"
+version = "2.26.0"
 description = "Olaf The Vibecoder — a voice-controlled coding assistant using OpenAI Realtime API + Claude Code"
 readme = "README.md"
 requires-python = ">=3.10"

{voice_vibecoder-2.24.1 → voice_vibecoder-2.26.0}/src/voice_vibecoder/__init__.py

@@ -3,7 +3,7 @@
 A voice-controlled coding assistant using OpenAI Realtime API + Claude Code.
 """
 
-__version__ = "2.24.1"
+__version__ = "2.26.0"
 
 from voice_vibecoder.app_config import VoiceConfig
 from voice_vibecoder.ui.screen import VoiceCodingScreen

{voice_vibecoder-2.24.1 → voice_vibecoder-2.26.0}/src/voice_vibecoder/config.py

@@ -30,7 +30,8 @@ def _agent_names(enabled: list[str] | None = None) -> str:
 
 _SYSTEM_INSTRUCTIONS: dict[Language, str] = {
     "no": """\
-Du er Olaf, en hjelpsom stemmeassistent for utviklere. Snakk alltid norsk. \
+Du er Olaf, en hjelpsom stemmeassistent for utviklere. Snakk norsk som standard, \
+men bytt språk umiddelbart hvis brukeren snakker eller ber om et annet språk. \
 Brukeren skal oppleve en helt vanlig samtale — du bare hjelper. \
 Dette er en talesamtale: svar kort, én til to setninger, med mindre brukeren ber om mer.
 
@@ -76,6 +77,10 @@ Avbryt/reset:
 - "start på nytt", "reset" → reset_agent
 - remove_branch_instance / remove_session for å fjerne instanser.
 
+Merge:
+- "merge det", "ta det inn i main" → merge_branch. Ved konflikter: foreslå å be \
+agenten merge main inn i sin branch først.
+
 Visning:
 - "vis diff" → show_diff
 - "vis output", "gå tilbake" → show_output
@@ -84,7 +89,8 @@ Visning:
 Spørsmål fra agenten:
 - Still spørsmålet til brukeren. Videresend svaret ordrett via answer_agent_question.""",
     "en": """\
-You are Olaf, a helpful voice assistant for developers. Always speak English. \
+You are Olaf, a helpful voice assistant for developers. Speak English by default, \
+but switch immediately if the user speaks or asks for another language. \
 The user should experience a normal conversation — you just help. \
 This is a spoken conversation: keep answers short, one or two sentences, unless the user asks for more.
 
@@ -130,6 +136,10 @@ Cancel/reset:
 - "start fresh", "reset" → reset_agent
 - remove_branch_instance / remove_session to remove instances.
 
+Merge:
+- "merge it", "bring it into main" → merge_branch. On conflicts: suggest asking \
+the agent to merge main into its branch first.
+
 Views:
 - "show diff" → show_diff
 - "show output", "go back" → show_output
@@ -138,7 +148,8 @@ Views:
 Questions from the agent:
 - Ask the user the question. Forward their answer verbatim via answer_agent_question.""",
     "sv": """\
-Du är Olaf, en hjälpsam röstassistent för utvecklare. Tala alltid svenska. \
+Du är Olaf, en hjälpsam röstassistent för utvecklare. Tala svenska som standard, \
+men byt språk direkt om användaren talar eller ber om ett annat språk. \
 Användaren ska uppleva ett helt vanligt samtal — du bara hjälper. \
 Detta är ett röstsamtal: svara kort, en till två meningar, om inte användaren ber om mer.
 
@@ -184,6 +195,10 @@ Avbryt/reset:
 - "börja om", "reset" → reset_agent
 - remove_branch_instance / remove_session för att ta bort instanser.
 
+Merge:
+- "merga det", "ta in det i main" → merge_branch. Vid konflikter: föreslå att \
+agenten mergar main in i sin branch först.
+
 Visning:
 - "visa diff" → show_diff
 - "visa output", "gå tillbaka" → show_output

{voice_vibecoder-2.24.1 → voice_vibecoder-2.26.0}/src/voice_vibecoder/session.py

@@ -608,7 +608,7 @@ class RealtimeSession:
     # -- Function call handling --
 
     # Tools that might do slow I/O (git subprocess, diff generation)
-    _SLOW_TOOLS = {"show_diff", "list_branches", "delete_worktree"}
+    _SLOW_TOOLS = {"show_diff", "list_branches", "delete_worktree", "merge_branch"}
 
     async def _handle_function_call(self, call_id: str, name: str, arguments: dict) -> None:
         if self._on_tool_call:

{voice_vibecoder-2.24.1 → voice_vibecoder-2.26.0}/src/voice_vibecoder/tools/definitions.py

@@ -332,6 +332,26 @@ _CORE_TOOLS: list[dict[str, Any]] = [
             "required": ["session_name"],
         },
     },
+    {
+        "type": "function",
+        "name": "merge_branch",
+        "description": (
+            "Merge a branch's work into the main branch. Commits any pending "
+            "work in the branch's worktree first. Use when the user approves "
+            "the changes and wants them merged (e.g. 'merge it', 'ta det inn "
+            "i main'). Reports conflicts without merging if any."
+        ),
+        "parameters": {
+            "type": "object",
+            "properties": {
+                "branch": {
+                    "type": "string",
+                    "description": "The branch to merge into main. Fuzzy-matched.",
+                },
+            },
+            "required": ["branch"],
+        },
+    },
     {
         "type": "function",
         "name": "delete_worktree",

{voice_vibecoder-2.24.1 → voice_vibecoder-2.26.0}/src/voice_vibecoder/tools/dispatch.py

@@ -199,6 +199,7 @@ def handle_tool_call(name: str, arguments: dict[str, Any]) -> str:
         _handle_delete_worktree,
         _handle_get_agent_status,
         _handle_list_branches,
+        _handle_merge_branch,
         _handle_remove_branch_instance,
         _handle_remove_session,
         _handle_reset_agent,
@@ -228,6 +229,8 @@ def handle_tool_call(name: str, arguments: dict[str, Any]) -> str:
         )
     elif name == "create_branch_instance":
         return _handle_create_branch_instance(arguments.get("branch", ""))
+    elif name == "merge_branch":
+        return _handle_merge_branch(arguments.get("branch", ""))
     elif name == "get_agent_status":
         return _handle_get_agent_status(arguments.get("branch"), arguments.get("session_name"))
     elif name == "cancel_agent":

{voice_vibecoder-2.24.1 → voice_vibecoder-2.26.0}/src/voice_vibecoder/tools/handlers.py

@@ -70,6 +70,20 @@ def get_startup_context() -> str:
             for wt in available:
                 parts.append(f"  - {wt.branch}")
 
+    if _dispatch._registry:
+        recent = _dispatch._registry.get_transcripts()[-8:]
+        if recent:
+            parts.append(
+                "This is a RESUMED conversation. Recent history (oldest first) — "
+                "continue naturally from here, do not greet as if meeting for the "
+                "first time:"
+            )
+            for entry in recent:
+                role = entry.get("role", "user")
+                text = entry.get("text", "").strip().replace("\n", " ")[:200]
+                if text:
+                    parts.append(f"  {role}: {text}")
+
     if not parts:
         return ""
     return "Current state:\n" + "\n".join(parts)
@@ -481,6 +495,35 @@ def _handle_remove_session(session_name: str) -> str:
     return f"Removed session '{instance.display_name}'."
 
 
+def _handle_merge_branch(branch: str) -> str:
+    if not branch.strip():
+        return "Error: branch name is required."
+    if not _dispatch._repo_root:
+        return "Error: repo root not configured."
+
+    from voice_vibecoder.tools.dispatch import _fuzzy_match_branch
+    from voice_vibecoder.worktrees import list_worktrees, merge_branch
+
+    worktrees = list_worktrees(_dispatch._repo_root)
+    wt_branches = [w.branch for w in worktrees if w.branch and not w.is_main]
+    resolved = _fuzzy_match_branch(branch, wt_branches) or branch
+
+    result = merge_branch(resolved, _dispatch._repo_root)
+    if result.merged:
+        return (
+            f"{result.message} The worktree is still there if more work is "
+            "needed — say so if it should be cleaned up (delete_worktree)."
+        )
+    if result.conflict_files:
+        files = ", ".join(result.conflict_files[:5])
+        return (
+            f"{result.message} Conflicting files: {files}. "
+            "Suggest sending the agent a request to resolve the conflicts by "
+            "merging main into its branch first."
+        )
+    return result.message
+
+
 def _handle_delete_worktree(branch: str, delete_branch: bool = False) -> str:
     if not branch.strip():
         return "Error: branch name is required."

{voice_vibecoder-2.24.1 → voice_vibecoder-2.26.0}/src/voice_vibecoder/voice_providers/__init__.py

@@ -144,6 +144,7 @@ def create_provider(settings) -> VoiceProvider:
         ws_url=settings.ws_url,
         ws_headers=settings.ws_headers,
         transcription_model="whisper-1" if is_azure else "gpt-4o-mini-transcribe",
+        transcription_language=settings.language,
         # Azure realtime is the GA API (/openai/v1/realtime); it uses the nested
         # session schema and response.output_audio.* events.
         ga=is_azure,

{voice_vibecoder-2.24.1 → voice_vibecoder-2.26.0}/src/voice_vibecoder/voice_providers/openai.py

@@ -41,11 +41,13 @@ class OpenAIVoiceProvider:
         ws_url: str,
         ws_headers: dict[str, str],
         transcription_model: str = "gpt-4o-mini-transcribe",
+        transcription_language: str = "",
         ga: bool = False,
     ) -> None:
         self._ws_url = ws_url
         self._ws_headers = ws_headers
         self._transcription_model = transcription_model
+        self._transcription_language = transcription_language
         # GA realtime API (Azure /openai/v1/realtime): nested `audio` session
         # schema + response.output_audio.* events. Preview otherwise.
         self._ga = ga
@@ -76,7 +78,10 @@ class OpenAIVoiceProvider:
                 "turn_detection": self._turn_detection(vad),
             }
             if self._transcription_model:
-                audio_input["transcription"] = {"model": self._transcription_model}
+                transcription = {"model": self._transcription_model}
+                if self._transcription_language:
+                    transcription["language"] = self._transcription_language
+                audio_input["transcription"] = transcription
             session: dict[str, Any] = {
                 "type": "realtime",
                 "output_modalities": ["audio"],
@@ -99,7 +104,11 @@ class OpenAIVoiceProvider:
             "voice": voice,
             "input_audio_format": "pcm16",
             "output_audio_format": "pcm16",
-            "input_audio_transcription": {"model": self._transcription_model},
+            "input_audio_transcription": (
+                {"model": self._transcription_model, "language": self._transcription_language}
+                if self._transcription_language
+                else {"model": self._transcription_model}
+            ),
             "tools": tools,
             "turn_detection": self._turn_detection(vad),
         }

{voice_vibecoder-2.24.1 → voice_vibecoder-2.26.0}/src/voice_vibecoder/worktrees.py

@@ -22,6 +22,12 @@ def configure_paths(worktrees_dir: Path) -> None:
     WORKTREES_DIR = worktrees_dir
 
 
+def _git(args: list[str], cwd: str) -> subprocess.CompletedProcess:
+    return subprocess.run(
+        ["git", *args], cwd=cwd, capture_output=True, text=True, timeout=30,
+    )
+
+
 @dataclass
 class WorktreeInfo:
     path: Path
@@ -191,6 +197,73 @@ def remove_worktree(branch: str, repo_root: Path, delete_branch: bool = False) -
         )
 
 
+@dataclass
+class MergeResult:
+    merged: bool
+    message: str
+    conflict_files: list[str] | None = None
+
+
+def merge_branch(branch: str, repo_root: Path) -> MergeResult:
+    """Merge a worktree branch into the branch checked out at repo_root.
+
+    Commits any uncommitted work in the branch's worktree first (agents
+    often leave files uncommitted), then merges. On conflict the merge is
+    aborted and the conflicting files are reported.
+    """
+    worktrees = list_worktrees(repo_root)
+    wt = next((w for w in worktrees if w.branch == branch), None)
+    if not wt:
+        return MergeResult(False, f"No worktree found for branch '{branch}'.")
+    if wt.is_main:
+        return MergeResult(False, "That is the main worktree — nothing to merge.")
+    if not wt.path.exists():
+        return MergeResult(False, f"Worktree directory for '{branch}' is missing.")
+
+    status = _git(["status", "--porcelain"], str(wt.path))
+    if status.returncode == 0 and status.stdout.strip():
+        _git(["add", "-A"], str(wt.path))
+        commit = _git(
+            ["commit", "-m", f"chore: voice-agent work on {branch}"], str(wt.path)
+        )
+        if commit.returncode != 0:
+            return MergeResult(
+                False, f"Could not commit pending work on '{branch}': "
+                f"{(commit.stderr or commit.stdout).strip()[:200]}"
+            )
+
+    target = get_current_branch(repo_root) or "main"
+    merge = _git(
+        ["merge", "--no-ff", branch, "-m", f"Merge branch '{branch}'"],
+        str(repo_root),
+    )
+    if merge.returncode != 0:
+        conflicts = _git(
+            ["diff", "--name-only", "--diff-filter=U"], str(repo_root)
+        )
+        conflict_files = [
+            f for f in conflicts.stdout.strip().splitlines() if f.strip()
+        ] if conflicts.returncode == 0 else []
+        _git(["merge", "--abort"], str(repo_root))
+        if conflict_files:
+            return MergeResult(
+                False,
+                f"Merge of '{branch}' into {target} has conflicts — aborted.",
+                conflict_files,
+            )
+        return MergeResult(
+            False,
+            f"Merge of '{branch}' failed: {(merge.stderr or merge.stdout).strip()[:200]}",
+        )
+
+    stat = _git(["diff", "--stat", "HEAD~1", "HEAD"], str(repo_root))
+    summary = stat.stdout.strip().splitlines()[-1].strip() if stat.returncode == 0 and stat.stdout.strip() else ""
+    msg = f"Merged '{branch}' into {target}."
+    if summary:
+        msg += f" {summary}."
+    return MergeResult(True, msg)
+
+
 def resolve_worktree_path(branch: str, repo_root: Path) -> Path:
     """Ensure a worktree exists for the branch.
 

voice_vibecoder-2.26.0/tests/test_merge_branch.py

@@ -0,0 +1,78 @@
+"""Voice-driven merge: merge a worktree branch into main, handle conflicts."""
+
+import subprocess
+from pathlib import Path
+
+import pytest
+
+from voice_vibecoder import worktrees
+from voice_vibecoder.worktrees import ensure_worktree, merge_branch
+
+
+def run(cwd: Path, *args: str) -> subprocess.CompletedProcess:
+    return subprocess.run(
+        args, cwd=cwd, capture_output=True, text=True, check=True
+    )
+
+
+@pytest.fixture
+def repo(tmp_path):
+    repo = tmp_path / "repo"
+    repo.mkdir()
+    run(repo, "git", "init", "-q", "-b", "main")
+    run(repo, "git", "config", "user.email", "t@t")
+    run(repo, "git", "config", "user.name", "t")
+    (repo / "README.md").write_text("hello\n")
+    run(repo, "git", "add", "README.md")
+    run(repo, "git", "commit", "-q", "-m", "init")
+    worktrees.configure_paths(tmp_path)
+    worktrees.invalidate_worktree_cache()
+    yield repo
+    worktrees.invalidate_worktree_cache()
+
+
+def test_merge_committed_work(repo):
+    wt = ensure_worktree("poem", repo)
+    (wt / "dikt.txt").write_text("roser er røde\n")
+    run(wt, "git", "add", "dikt.txt")
+    run(wt, "git", "commit", "-q", "-m", "add poem")
+    worktrees.invalidate_worktree_cache()
+
+    result = merge_branch("poem", repo)
+    assert result.merged, result.message
+    assert (repo / "dikt.txt").exists()
+
+
+def test_merge_commits_pending_work_first(repo):
+    wt = ensure_worktree("wip", repo)
+    (wt / "uncommitted.txt").write_text("agents forget to commit\n")
+    worktrees.invalidate_worktree_cache()
+
+    result = merge_branch("wip", repo)
+    assert result.merged, result.message
+    assert (repo / "uncommitted.txt").exists()
+
+
+def test_merge_conflict_aborts_and_reports(repo):
+    wt = ensure_worktree("clash", repo)
+    (wt / "README.md").write_text("branch version\n")
+    run(wt, "git", "add", "README.md")
+    run(wt, "git", "commit", "-q", "-m", "branch edit")
+    (repo / "README.md").write_text("main version\n")
+    run(repo, "git", "add", "README.md")
+    run(repo, "git", "commit", "-q", "-m", "main edit")
+    worktrees.invalidate_worktree_cache()
+
+    result = merge_branch("clash", repo)
+    assert not result.merged
+    assert result.conflict_files == ["README.md"]
+    # main is left clean — merge was aborted
+    status = run(repo, "git", "status", "--porcelain").stdout.strip()
+    assert status == ""
+    assert (repo / "README.md").read_text() == "main version\n"
+
+
+def test_merge_unknown_branch(repo):
+    result = merge_branch("nope", repo)
+    assert not result.merged
+    assert "No worktree" in result.message

voice_vibecoder-2.26.0/tests/test_startup_context.py

@@ -0,0 +1,28 @@
+"""Resumed sessions must carry recent conversation into the system prompt."""
+
+from voice_vibecoder.instances import InstanceRegistry
+from voice_vibecoder.tools import configure
+from voice_vibecoder.tools.handlers import get_startup_context
+
+
+def make_registry() -> InstanceRegistry:
+    reg = InstanceRegistry()
+    reg._transcript_buffer = [
+        {"role": "user", "text": "lag en fil i repoet"},
+        {"role": "assistant", "text": "Klart, jeg setter i gang!"},
+    ]
+    return reg
+
+
+def test_startup_context_includes_recent_transcript():
+    configure(registry=make_registry(), repo_root=None)
+    ctx = get_startup_context()
+    assert "RESUMED" in ctx
+    assert "lag en fil i repoet" in ctx
+    assert "Klart, jeg setter i gang!" in ctx
+
+
+def test_startup_context_empty_without_history():
+    configure(registry=InstanceRegistry(), repo_root=None)
+    ctx = get_startup_context()
+    assert "RESUMED" not in ctx