diff --git a/backend/app/contracts/__init__.py b/backend/app/contracts/__init__.py
index 0aabd6fd..37ecf0e2 100644
--- a/backend/app/contracts/__init__.py
+++ b/backend/app/contracts/__init__.py
@@ -68,6 +68,19 @@
     ApiKeyStatus,
     ApiKeysListResponse,
 )
+from .provider_types import (
+    ALL_PROVIDER_TYPES,
+    LEGACY_GEMINI,
+    PROVIDER_ANTHROPIC,
+    PROVIDER_CUSTOM,
+    PROVIDER_GITHUB_COPILOT,
+    PROVIDER_GOOGLE,
+    PROVIDER_OLLAMA,
+    PROVIDER_OLLAMA_CLOUD,
+    PROVIDER_OPENAI,
+    PROVIDER_OPENAI_COMPATIBLE,
+    ProviderType,
+)
 from .report_v3 import (
     Claim,
     ChangeRecommendation,
@@ -136,6 +149,17 @@
     "RunSummary",
     "SegmentReach",
     "VoiceRegister",
+    "ALL_PROVIDER_TYPES",
+    "LEGACY_GEMINI",
+    "PROVIDER_ANTHROPIC",
+    "PROVIDER_CUSTOM",
+    "PROVIDER_GITHUB_COPILOT",
+    "PROVIDER_GOOGLE",
+    "PROVIDER_OLLAMA",
+    "PROVIDER_OLLAMA_CLOUD",
+    "PROVIDER_OPENAI",
+    "PROVIDER_OPENAI_COMPATIBLE",
+    "ProviderType",
     # ReportV3 — 11 Pflichtabschnitt-DTOs
     "Claim",
     "ChangeRecommendation",
diff --git a/backend/app/contracts/llm_profile_contract.py b/backend/app/contracts/llm_profile_contract.py
index e087ac84..e9ebde4e 100644
--- a/backend/app/contracts/llm_profile_contract.py
+++ b/backend/app/contracts/llm_profile_contract.py
@@ -6,13 +6,13 @@
 from __future__ import annotations
 
 from datetime import datetime
-from typing import Literal, Optional
+from typing import Optional
 
 from pydantic import BaseModel, ConfigDict, Field
 
-_STRICT = ConfigDict(extra="forbid", populate_by_name=True)
+from .provider_types import ProviderType
 
-ProviderLiteral = Literal["ollama", "openai", "gemini", "anthropic", "custom"]
+_STRICT = ConfigDict(extra="forbid", populate_by_name=True)
 
 
 class LlmProfile(BaseModel):
@@ -20,7 +20,7 @@ class LlmProfile(BaseModel):
 
     id: str = Field(..., description="UUID, server-generiert")
     name: str = Field(..., min_length=1, max_length=80)
-    provider: ProviderLiteral
+    provider: ProviderType
     base_url: str = Field(..., min_length=1)
     model_name: str = Field(..., min_length=1)
     # None = nicht gesetzt (Update lässt Feld weg). "" = explizit geleert.
@@ -39,7 +39,7 @@ class LlmProfileCreateRequest(BaseModel):
     model_config = _STRICT
 
     name: str = Field(..., min_length=1, max_length=80)
-    provider: ProviderLiteral
+    provider: ProviderType
     base_url: str = Field(..., min_length=1)
     model_name: str = Field(..., min_length=1)
     # None = nicht gesetzt (Update lässt Feld weg). "" = explizit geleert.
diff --git a/backend/app/contracts/llm_routing_contract.py b/backend/app/contracts/llm_routing_contract.py
index 8f6571c8..29072489 100644
--- a/backend/app/contracts/llm_routing_contract.py
+++ b/backend/app/contracts/llm_routing_contract.py
@@ -6,6 +6,7 @@
 
 from typing import Dict, Literal, Optional, Any, List
 from pydantic import BaseModel, ConfigDict, Field, model_validator
+from .provider_types import ProviderType
 
 _STRICT = ConfigDict(extra="forbid")
 
@@ -71,7 +72,7 @@ class ProviderDescriptor(BaseModel):
 
     id: str
     label: str
-    type: Literal["ollama_cloud", "openai", "google", "openai_compatible", "github_copilot"]
+    type: ProviderType
     base_url: Optional[str] = None
     api_key_ref: Optional[str] = None
     supports_models_endpoint: bool = False
diff --git a/backend/app/contracts/provider_types.py b/backend/app/contracts/provider_types.py
new file mode 100644
index 00000000..0bfdee01
--- /dev/null
+++ b/backend/app/contracts/provider_types.py
@@ -0,0 +1,41 @@
+from typing import Literal
+
+PROVIDER_OLLAMA = "ollama"
+PROVIDER_OPENAI = "openai"
+PROVIDER_GOOGLE = "google"
+PROVIDER_ANTHROPIC = "anthropic"
+PROVIDER_CUSTOM = "custom"
+PROVIDER_OLLAMA_CLOUD = "ollama_cloud"
+PROVIDER_OPENAI_COMPATIBLE = "openai_compatible"
+PROVIDER_GITHUB_COPILOT = "github_copilot"
+PROVIDER_CLOUD = "cloud"
+PROVIDER_UNKNOWN = "unknown"
+
+# Legacy alias
+LEGACY_GEMINI = "gemini"
+
+ProviderType = Literal[
+    "ollama",
+    "openai",
+    "google",
+    "anthropic",
+    "custom",
+    "ollama_cloud",
+    "openai_compatible",
+    "github_copilot",
+    "cloud",
+    "unknown",
+]
+
+ALL_PROVIDER_TYPES: tuple[ProviderType, ...] = (
+    PROVIDER_OLLAMA,
+    PROVIDER_OPENAI,
+    PROVIDER_GOOGLE,
+    PROVIDER_ANTHROPIC,
+    PROVIDER_CUSTOM,
+    PROVIDER_OLLAMA_CLOUD,
+    PROVIDER_OPENAI_COMPATIBLE,
+    PROVIDER_GITHUB_COPILOT,
+    PROVIDER_CLOUD,
+    PROVIDER_UNKNOWN,
+)
diff --git a/backend/app/contracts/report_v3.py b/backend/app/contracts/report_v3.py
index 7e7bff90..e50b4c9e 100644
--- a/backend/app/contracts/report_v3.py
+++ b/backend/app/contracts/report_v3.py
@@ -15,6 +15,8 @@
 
 from pydantic import BaseModel, ConfigDict, Field
 
+from .provider_types import ProviderType
+
 
 _STRICT = ConfigDict(extra="forbid", str_strip_whitespace=True)
 
@@ -236,7 +238,7 @@ class ModelAttribution(BaseModel):
     model_config = _STRICT
 
     stage: ModelAttributionStage
-    provider: str = Field(min_length=1, description="z. B. 'ollama', 'openai', 'gemini'")
+    provider: ProviderType = Field(description="z. B. 'ollama', 'openai', 'google'")
     model_id: str = Field(min_length=1, description="Backend-Modell-ID, z. B. 'qwen2.5:32b'")
     prompt_tokens: int | None = Field(default=None, ge=0)
     completion_tokens: int | None = Field(default=None, ge=0)
diff --git a/backend/app/services/llm_profiles_store.py b/backend/app/services/llm_profiles_store.py
index 305a508c..939cbd36 100644
--- a/backend/app/services/llm_profiles_store.py
+++ b/backend/app/services/llm_profiles_store.py
@@ -14,7 +14,15 @@
 from pathlib import Path
 from typing import Optional
 
-from ..contracts.llm_profile_contract import LlmProfile, LlmProfileCreateRequest
+from ..contracts import (
+    PROVIDER_ANTHROPIC,
+    PROVIDER_CUSTOM,
+    PROVIDER_GOOGLE,
+    PROVIDER_OLLAMA,
+    PROVIDER_OPENAI,
+    LlmProfile,
+    LlmProfileCreateRequest,
+)
 
 
 def _now() -> datetime:
@@ -77,15 +85,15 @@ def _bootstrap_profile() -> Optional[dict]:
     if not model:
         return None
     if "openai.com" in base_url:
-        provider = "openai"
+        provider = PROVIDER_OPENAI
     elif "googleapis.com" in base_url:
-        provider = "gemini"
+        provider = PROVIDER_GOOGLE
     elif "anthropic.com" in base_url:
-        provider = "anthropic"
+        provider = PROVIDER_ANTHROPIC
     elif any(h in base_url for h in ("localhost", "127.0.0.1", "host.docker.internal")):
-        provider = "ollama"
+        provider = PROVIDER_OLLAMA
     else:
-        provider = "custom"
+        provider = PROVIDER_CUSTOM
     return dict(
         id=uuid.uuid4().hex,
         name="Standard",
diff --git a/backend/app/services/llm_provider_registry.py b/backend/app/services/llm_provider_registry.py
index 8b3f9137..6cb7c594 100644
--- a/backend/app/services/llm_provider_registry.py
+++ b/backend/app/services/llm_provider_registry.py
@@ -4,6 +4,13 @@
 """
 
 from typing import List, Optional
+from ..contracts import (
+    PROVIDER_GITHUB_COPILOT,
+    PROVIDER_GOOGLE,
+    PROVIDER_OLLAMA_CLOUD,
+    PROVIDER_OPENAI,
+    PROVIDER_OPENAI_COMPATIBLE,
+)
 from ..contracts.llm_routing_contract import ProviderDescriptor
 
 class LlmProviderRegistry:
@@ -20,7 +27,7 @@ def get_providers(self, session_api_keys: Optional[dict] = None) -> List[Provide
             ProviderDescriptor(
                 id="ollama_cloud",
                 label="Ollama (Cloud)",
-                type="ollama_cloud",
+                type=PROVIDER_OLLAMA_CLOUD,
                 # OpenAI-kompatibler Ollama-Cloud-Endpoint. Auth via Bearer-Token
                 # (``OLLAMA_API_KEY``). Doku: https://docs.ollama.com/cloud
                 base_url="https://ollama.com/v1",
@@ -36,7 +43,7 @@ def get_providers(self, session_api_keys: Optional[dict] = None) -> List[Provide
             ProviderDescriptor(
                 id="openai",
                 label="OpenAI",
-                type="openai",
+                type=PROVIDER_OPENAI,
                 base_url="https://api.openai.com/v1",
                 api_key_ref="OPENAI_API_KEY",
                 supports_models_endpoint=True,
@@ -45,7 +52,7 @@ def get_providers(self, session_api_keys: Optional[dict] = None) -> List[Provide
             ProviderDescriptor(
                 id="google",
                 label="Google Gemini",
-                type="google",
+                type=PROVIDER_GOOGLE,
                 base_url="https://generativelanguage.googleapis.com/v1beta/openai/",
                 api_key_ref="GOOGLE_API_KEY",
                 supports_models_endpoint=True,
@@ -54,7 +61,7 @@ def get_providers(self, session_api_keys: Optional[dict] = None) -> List[Provide
             ProviderDescriptor(
                 id="openai_compatible",
                 label="OpenAI Compatible",
-                type="openai_compatible",
+                type=PROVIDER_OPENAI_COMPATIBLE,
                 api_key_ref="LLM_API_KEY",
                 supports_models_endpoint=True,
                 fallback_models=[],
@@ -62,7 +69,7 @@ def get_providers(self, session_api_keys: Optional[dict] = None) -> List[Provide
             ProviderDescriptor(
                 id="github_copilot",
                 label="GitHub Copilot",
-                type="github_copilot",
+                type=PROVIDER_GITHUB_COPILOT,
                 base_url="https://api.githubcopilot.com",
                 api_key_ref="GH_AUTH_TOKEN",
                 supports_models_endpoint=False,
diff --git a/backend/app/services/llm_runtime.py b/backend/app/services/llm_runtime.py
index f1425d08..08157573 100644
--- a/backend/app/services/llm_runtime.py
+++ b/backend/app/services/llm_runtime.py
@@ -11,18 +11,25 @@
 from typing import Any, Mapping, Optional
 from urllib.parse import urlparse
 
+from ..contracts import (
+    LEGACY_GEMINI,
+    PROVIDER_CUSTOM,
+    PROVIDER_GOOGLE,
+    PROVIDER_OPENAI,
+    PROVIDER_OPENAI_COMPATIBLE,
+)
 
 PROVIDER_DEFAULT_BASE_URLS = {
-    "google": "https://generativelanguage.googleapis.com/v1beta/openai/",
-    "openai": "https://api.openai.com/v1",
+    PROVIDER_GOOGLE: "https://generativelanguage.googleapis.com/v1beta/openai/",
+    PROVIDER_OPENAI: "https://api.openai.com/v1",
 }
 
 _PROVIDER_ALIASES = {
     "default": "default",
     "server": "default",
-    "google": "google",
-    "gemini": "google",
-    "openai": "openai",
+    "google": PROVIDER_GOOGLE,
+    LEGACY_GEMINI: PROVIDER_GOOGLE,
+    "openai": PROVIDER_OPENAI,
     "custom": "custom_openai",
     "custom_openai": "custom_openai",
     "openai_compatible": "custom_openai",
@@ -33,8 +40,8 @@
 # beliebigem Format (``custom_openai``, ``ollama_cloud``, ``github_copilot``)
 # stehen NICHT drin — die werden nicht validiert.
 _KEY_PREFIX_BY_PROVIDER: dict[str, tuple[str, ...]] = {
-    "openai": ("sk-",),
-    "google": ("AIzaSy",),
+    PROVIDER_OPENAI: ("sk-",),
+    PROVIDER_GOOGLE: ("AIzaSy",),
 }
 
 
@@ -116,7 +123,7 @@ def parse_runtime_llm_config(data: Mapping[str, Any]) -> RuntimeLlmConfig:
     provider = _PROVIDER_ALIASES.get(provider_raw)
     if provider is None:
         raise ValueError(
-            "llm_provider.provider must be one of: default, google, openai, custom_openai"
+            f"llm_provider.provider must be one of: default, {PROVIDER_GOOGLE}, {PROVIDER_OPENAI}, custom_openai"
         )
     if provider == "default":
         return RuntimeLlmConfig()
diff --git a/backend/app/services/model_catalog_service.py b/backend/app/services/model_catalog_service.py
index 1ccbc1ba..2a84eccf 100644
--- a/backend/app/services/model_catalog_service.py
+++ b/backend/app/services/model_catalog_service.py
@@ -17,6 +17,14 @@
 import urllib3
 from pydantic import BaseModel, ConfigDict
 
+from ..contracts import (
+    PROVIDER_GOOGLE,
+    PROVIDER_OLLAMA_CLOUD,
+    PROVIDER_OPENAI,
+    PROVIDER_OPENAI_COMPATIBLE,
+    PROVIDER_GITHUB_COPILOT,
+)
+
 from ..utils.logger import get_logger
 
 logger = get_logger("agora.model_catalog")
@@ -122,11 +130,11 @@ def get_models(self, provider_id: str, provider_type: str, base_url: str, api_ke
 
     def _fetch_live(self, provider_type: str, base_url: str, api_key: Optional[str]) -> List[str]:
         """Discovery implementation per provider type."""
-        if provider_type == "github_copilot":
+        if provider_type == PROVIDER_GITHUB_COPILOT:
             # Phase 1: kein Live-Discovery — statische Liste über _get_fallbacks.
             return []
 
-        if provider_type == "ollama_cloud":
+        if provider_type == PROVIDER_OLLAMA_CLOUD:
             # Ollama Cloud (ollama.com) BRAUCHT Bearer-Token am /v1/models-Endpoint.
             # Vorher fehlte der Header — Ergebnis war ein stiller 401 → leere Liste
             # (PR #528 Follow-up).
@@ -143,7 +151,7 @@ def _fetch_live(self, provider_type: str, base_url: str, api_key: Optional[str])
                 return [m["name"] for m in data.get("models", [])]
             return []
 
-        if provider_type in ("openai", "google", "openai_compatible"):
+        if provider_type in (PROVIDER_OPENAI, PROVIDER_GOOGLE, PROVIDER_OPENAI_COMPATIBLE):
             # OpenAI-shape: data[].id. Erst /models, dann /v1/models als Fallback,
             # falls die base_url nicht schon /v1 enthält.
             data = _http_get_json(f"{base_url.rstrip('/')}/models", api_key=api_key)
@@ -161,13 +169,13 @@ def _get_fallbacks(self, provider_type: str) -> List[str]:
         # zwingend installiert (User-Bericht 2026-05-16: halluzinierte Einträge
         # im Dashboard-Picker). Lieber leeres Catalog + sichtbarer Fehlerzustand
         # als Ehrlichkeits-Lüge mit nicht-existenten Modellen.
-        if provider_type == "ollama_cloud":
+        if provider_type == PROVIDER_OLLAMA_CLOUD:
             return []
-        if provider_type == "openai":
+        if provider_type == PROVIDER_OPENAI:
             return ["gpt-4o", "gpt-4o-mini", "o1-preview"]
-        if provider_type == "google":
+        if provider_type == PROVIDER_GOOGLE:
             return ["gemini-1.5-pro", "gemini-1.5-flash"]
-        if provider_type == "github_copilot":
+        if provider_type == PROVIDER_GITHUB_COPILOT:
             from .llm_providers.github_copilot import GITHUB_COPILOT_MODELS
             return list(GITHUB_COPILOT_MODELS)
         return []
diff --git a/backend/app/services/model_event_bus.py b/backend/app/services/model_event_bus.py
index 368a6012..6c6b19c5 100644
--- a/backend/app/services/model_event_bus.py
+++ b/backend/app/services/model_event_bus.py
@@ -28,6 +28,7 @@
 from typing import Any, Generator, Iterator, Literal
 
 from pydantic import BaseModel, ConfigDict
+from ..contracts import ProviderType
 
 from ..utils.logger import get_logger
 
@@ -61,16 +62,7 @@ class ModelActiveEvent(BaseModel):
         "graph",
         "unknown",
     ]
-    provider: Literal[
-        "ollama",
-        "cloud",
-        "openai",
-        "google",
-        "ollama_cloud",
-        "openai_compatible",
-        "github_copilot",
-        "unknown",
-    ]
+    provider: ProviderType
     ts: float
     extra: dict[str, Any] | None = None
 
diff --git a/backend/app/services/runtime_run_config.py b/backend/app/services/runtime_run_config.py
index c5ce0635..bd2e015c 100644
--- a/backend/app/services/runtime_run_config.py
+++ b/backend/app/services/runtime_run_config.py
@@ -8,6 +8,12 @@
 import tempfile
 from urllib.parse import urlparse, urlunparse
 from typing import Optional, Dict, Any
+from ..contracts import (
+    PROVIDER_GOOGLE,
+    PROVIDER_OLLAMA_CLOUD,
+    PROVIDER_OPENAI,
+    PROVIDER_OPENAI_COMPATIBLE,
+)
 from ..contracts.llm_routing_contract import RuntimeLlmRouting, StageLLMRoute, StageId
 from ..utils.artifact_locator import ArtifactLocator
 from ..utils.logger import get_logger
@@ -47,12 +53,12 @@ def _detect_default_provider_id(base_url: Optional[str], model_name: Optional[st
     hostname = (parsed.hostname or "").lower() if parsed else ""
 
     if normalized_model.endswith(":cloud") or hostname == "ollama.com":
-        return "ollama_cloud"
+        return PROVIDER_OLLAMA_CLOUD
     if hostname == "generativelanguage.googleapis.com" or "gemini" in normalized_model:
-        return "google"
+        return PROVIDER_GOOGLE
     if hostname in {"api.openai.com", "openai.com"}:
-        return "openai"
-    return "openai_compatible"
+        return PROVIDER_OPENAI
+    return PROVIDER_OPENAI_COMPATIBLE
 
 
 class RuntimeRunConfig:
diff --git a/backend/app/services/secret_resolver.py b/backend/app/services/secret_resolver.py
index 689c63d5..2f883345 100644
--- a/backend/app/services/secret_resolver.py
+++ b/backend/app/services/secret_resolver.py
@@ -38,6 +38,13 @@
 from typing import Optional, Dict
 
 from ..config import Config
+from ..contracts import (
+    PROVIDER_GOOGLE,
+    PROVIDER_OLLAMA_CLOUD,
+    PROVIDER_OPENAI,
+    PROVIDER_OPENAI_COMPATIBLE,
+    PROVIDER_GITHUB_COPILOT,
+)
 from .llm_provider_secrets_store import get_llm_provider_secrets_store
 
 logger = logging.getLogger("agora.secret_resolver")
@@ -81,9 +88,9 @@ def _format_valid(value: Optional[str], provider_type: str) -> bool:
         return False
     if v.lower() in _TOXIC_KEY_LITERALS:
         return False
-    if provider_type == "openai":
+    if provider_type == PROVIDER_OPENAI:
         return bool(_OPENAI_KEY_RE.match(v))
-    if provider_type == "google":
+    if provider_type == PROVIDER_GOOGLE:
         return bool(_GOOGLE_KEY_RE.match(v))
     return True
 
@@ -136,16 +143,16 @@ def get_api_key(self, provider_id: str, provider_type: str) -> Optional[str]:
 
         # 3. Provider-spezifische Environment-Variablen (mit Format-Sanity-Check)
         env_candidates: list[tuple[str, Optional[str]]] = []
-        if provider_type == "openai":
+        if provider_type == PROVIDER_OPENAI:
             env_candidates.append(("env:OPENAI_API_KEY", os.environ.get("OPENAI_API_KEY")))
-        elif provider_type == "google":
+        elif provider_type == PROVIDER_GOOGLE:
             env_candidates.append(("env:GOOGLE_API_KEY", os.environ.get("GOOGLE_API_KEY")))
             env_candidates.append(("env:GEMINI_API_KEY", os.environ.get("GEMINI_API_KEY")))
-        elif provider_type == "ollama_cloud":
+        elif provider_type == PROVIDER_OLLAMA_CLOUD:
             env_candidates.append(("env:OLLAMA_API_KEY", os.environ.get("OLLAMA_API_KEY")))
-        elif provider_type == "openai_compatible":
+        elif provider_type == PROVIDER_OPENAI_COMPATIBLE:
             env_candidates.append(("env:LLM_API_KEY", os.environ.get("LLM_API_KEY")))
-        elif provider_type == "github_copilot":
+        elif provider_type == PROVIDER_GITHUB_COPILOT:
             # GitHub Copilot: Token-Auflösung über separates Modul (Slice B).
             try:
                 from .llm_providers.github_copilot import resolve_copilot_token
@@ -174,7 +181,7 @@ def get_api_key(self, provider_id: str, provider_type: str) -> Optional[str]:
         # 4. Globaler Fallback Config.LLM_API_KEY
         fallback = Config.LLM_API_KEY
         if fallback:
-            if provider_type in ("openai", "google"):
+            if provider_type in (PROVIDER_OPENAI, PROVIDER_GOOGLE):
                 if _format_valid(fallback, provider_type):
                     self.last_source = "config_fallback"
                     return fallback
@@ -184,7 +191,7 @@ def get_api_key(self, provider_id: str, provider_type: str) -> Optional[str]:
                     "im UI (Settings → LLM-Provider) oder via ENV %s.",
                     provider_type,
                     _mask_for_log(fallback),
-                    "OPENAI_API_KEY" if provider_type == "openai" else "GOOGLE_API_KEY",
+                    "OPENAI_API_KEY" if provider_type == PROVIDER_OPENAI else "GOOGLE_API_KEY",
                 )
                 return None
             # Provider-Type ohne striktes Format (ollama_cloud, openai_compatible, …):
diff --git a/backend/app/utils/llm_profile_resolver.py b/backend/app/utils/llm_profile_resolver.py
index faac3f13..88fe3da5 100644
--- a/backend/app/utils/llm_profile_resolver.py
+++ b/backend/app/utils/llm_profile_resolver.py
@@ -16,13 +16,14 @@
 
 from typing import Any, Mapping
 
+from ..contracts import LEGACY_GEMINI, PROVIDER_GOOGLE
 from ..services.llm_profiles_store import get_llm_profiles_store
 
 _PROFILE_PREFIX = "profile:"
 
 _PROFILE_PROVIDER_TO_RUNTIME = {
     "openai": "openai",
-    "gemini": "google",
+    LEGACY_GEMINI: PROVIDER_GOOGLE,
     "ollama": "custom_openai",
     "anthropic": "custom_openai",
     "custom": "custom_openai",
diff --git a/backend/tests/contracts/test_provider_types.py b/backend/tests/contracts/test_provider_types.py
new file mode 100644
index 00000000..854244cf
--- /dev/null
+++ b/backend/tests/contracts/test_provider_types.py
@@ -0,0 +1,58 @@
+import re
+from pathlib import Path
+
+def test_no_gemini_literals_in_code():
+    """
+    Regression gate: Ensure 'gemini' literal is not used for provider identification
+    outside of comments, docstrings, and the provider_types.py mapping.
+    """
+    app_root = Path(__file__).parents[3] / "backend" / "app"
+
+    # Regex to find "gemini" (with quotes)
+    gemini_re = re.compile(r'"gemini"')
+
+    forbidden_lines = []
+
+    # Files/directories to skip completely
+    skip_files = {"provider_types.py"}
+
+    # Legitimate substrings to allow on a line containing "gemini"
+    legitimate_substrings = [
+        "fallback_models",
+        "gemini-1.5",
+        "gemini-3",
+        "VISION_MODEL_NAME",
+        "normalized_model",
+        "LEGACY_GEMINI"
+    ]
+
+    for path in app_root.rglob("*.py"):
+        if path.name in skip_files:
+            continue
+
+        with open(path, "r", encoding="utf-8") as f:
+            lines = f.readlines()
+
+        for i, line in enumerate(lines, 1):
+            if gemini_re.search(line):
+                # Check for comments: if # appears before "gemini"
+                comment_idx = line.find("#")
+                gemini_idx = line.find('"gemini"')
+                if comment_idx != -1 and comment_idx < gemini_idx:
+                    continue
+
+                # Rough docstring check
+                if '"""' in line or "'''" in line:
+                    continue
+
+                # Check for legitimate usage
+                if any(sub in line for sub in legitimate_substrings):
+                    continue
+
+                forbidden_lines.append(f"{path.relative_to(app_root.parent)}:{i}: {line.strip()}")
+
+    if forbidden_lines:
+        print("\nFound forbidden 'gemini' literals:")
+        for fl in forbidden_lines:
+            print(fl)
+        assert not forbidden_lines, f"Found {len(forbidden_lines)} forbidden 'gemini' literals."
diff --git a/frontend/src/contracts/llmProfileContract.ts b/frontend/src/contracts/llmProfileContract.ts
index 3888f953..7b8b6f1a 100644
--- a/frontend/src/contracts/llmProfileContract.ts
+++ b/frontend/src/contracts/llmProfileContract.ts
@@ -7,9 +7,15 @@ import { z } from "zod";
 export const LlmProviderSchema = z.enum([
   "ollama",
   "openai",
+  "google",
   "gemini",
   "anthropic",
   "custom",
+  "ollama_cloud",
+  "openai_compatible",
+  "github_copilot",
+  "cloud",
+  "unknown",
 ]);
 export type LlmProvider = z.infer<typeof LlmProviderSchema>;
 
diff --git a/schemas/llm-profile-create-request.schema.json b/schemas/llm-profile-create-request.schema.json
index 0dd2bb1d..33e52846 100644
--- a/schemas/llm-profile-create-request.schema.json
+++ b/schemas/llm-profile-create-request.schema.json
@@ -40,9 +40,14 @@
       "enum": [
         "ollama",
         "openai",
-        "gemini",
+        "google",
         "anthropic",
-        "custom"
+        "custom",
+        "ollama_cloud",
+        "openai_compatible",
+        "github_copilot",
+        "cloud",
+        "unknown"
       ],
       "title": "Provider",
       "type": "string"
diff --git a/schemas/llm-profile-list-response.schema.json b/schemas/llm-profile-list-response.schema.json
index f4593e97..b29c10e4 100644
--- a/schemas/llm-profile-list-response.schema.json
+++ b/schemas/llm-profile-list-response.schema.json
@@ -50,9 +50,14 @@
           "enum": [
             "ollama",
             "openai",
-            "gemini",
+            "google",
             "anthropic",
-            "custom"
+            "custom",
+            "ollama_cloud",
+            "openai_compatible",
+            "github_copilot",
+            "cloud",
+            "unknown"
           ],
           "title": "Provider",
           "type": "string"
diff --git a/schemas/llm-profile.schema.json b/schemas/llm-profile.schema.json
index 2c5c2753..6cf85df7 100644
--- a/schemas/llm-profile.schema.json
+++ b/schemas/llm-profile.schema.json
@@ -50,9 +50,14 @@
       "enum": [
         "ollama",
         "openai",
-        "gemini",
+        "google",
         "anthropic",
-        "custom"
+        "custom",
+        "ollama_cloud",
+        "openai_compatible",
+        "github_copilot",
+        "cloud",
+        "unknown"
       ],
       "title": "Provider",
       "type": "string"
diff --git a/schemas/llm-provider-descriptor.schema.json b/schemas/llm-provider-descriptor.schema.json
index dcc026f6..460379b3 100644
--- a/schemas/llm-provider-descriptor.schema.json
+++ b/schemas/llm-provider-descriptor.schema.json
@@ -50,11 +50,16 @@
     },
     "type": {
       "enum": [
-        "ollama_cloud",
+        "ollama",
         "openai",
         "google",
+        "anthropic",
+        "custom",
+        "ollama_cloud",
         "openai_compatible",
-        "github_copilot"
+        "github_copilot",
+        "cloud",
+        "unknown"
       ],
       "title": "Type",
       "type": "string"
diff --git a/schemas/report-v3.schema.json b/schemas/report-v3.schema.json
index ccdd3533..20ebcf5b 100644
--- a/schemas/report-v3.schema.json
+++ b/schemas/report-v3.schema.json
@@ -369,8 +369,19 @@
           "title": "Prompt Tokens"
         },
         "provider": {
-          "description": "z. B. 'ollama', 'openai', 'gemini'",
-          "minLength": 1,
+          "description": "z. B. 'ollama', 'openai', 'google'",
+          "enum": [
+            "ollama",
+            "openai",
+            "google",
+            "anthropic",
+            "custom",
+            "ollama_cloud",
+            "openai_compatible",
+            "github_copilot",
+            "cloud",
+            "unknown"
+          ],
           "title": "Provider",
           "type": "string"
         },