terradev-cli 5.4.1__tar.gz → 5.4.3__tar.gz

{terradev_cli-5.4.1 → terradev_cli-5.4.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: terradev-cli
-Version: 5.4.1
+Version: 5.4.3
 Summary: An imperative command-line-interface for AI workload orchestration.
 Author-email: Terradev Team <team@terradev.com>
 Maintainer-email: Terradev Team <team@terradev.com>
@@ -93,7 +93,7 @@ Requires-Dist: redis>=5.0.0; extra == "enterprise"
 Requires-Dist: PyJWT>=2.8.0; extra == "enterprise"
 Dynamic: license-file
 
-# Terradev CLI v5.3.8
+# Terradev CLI v5.4.3
 
 **An imperative command-line-interface for AI workload orchestration.**
 
@@ -111,6 +111,29 @@ Continued focus on lower cost, faster provisioning, and topology-aware execution
 
 Model agnostic. Dataset agnostic. GPU agnostic. Provider agnostic. The only thing Terradev is not agnostic about is correctness: it enforces topology, idempotency, and sequencing.
 
+**NOTES ON 5.4.3**
+
+Fixed **credential format fallback bug** in `_provider_creds()`:
+
+- **Bug**: When credentials file contains both nested format `{"runpod": {"api_key": "..."}}` and flat format `{"runpod_api_key": "..."}`, the `_provider_creds()` method prioritized nested format even if it contained placeholder values like "PASTE_NEW_RUNPOD_KEY_HERE". This caused `terradev quote` to fail because it used placeholder credentials instead of real flat-format credentials.
+
+- **Fix**: Added placeholder detection in `_provider_creds()` to check if nested values are placeholders (containing "paste_", "your_", "example_", "test_", "placeholder_", "xxx"). If nested format contains only placeholders, the method now falls back to flat format credentials.
+
+- **Impact**: Users with mixed credential formats (nested placeholders from configure + real flat keys from onboarding) now get real credentials from the flat format, enabling `terradev quote` to return actual pricing data.
+
+**NOTES ON 5.4.2**
+
+Fixed **credential type checking bug** in `is_first_time_user()` and onboarding:
+
+- **Bug**: `is_first_time_user()` and onboarding code called `.lower()` on credential values without checking if they were strings. The credentials file uses nested dict format `{"runpod": {"api_key": "..."}}`, so when iterating over values, some are dicts, not strings. This caused `AttributeError: 'dict' object has no attribute 'lower'`.
+
+- **Fix**:
+  - Added `isinstance(value, str)` check before calling `.lower()` in `is_first_time_user()` (line 119)
+  - Added `isinstance(existing_value, str)` check before calling `.lower()` in onboarding (line 764)
+  - The `_provider_creds()` method already correctly handles both flat and nested credential formats
+
+- **Impact**: `terradev quote` and other commands that check credentials now work correctly with the nested dict credential format.
+
 **NOTES ON 5.3.9**
 
 Added **LoRAX (LoRA eXchange) integration** and **HuggingFace PEFT import** for production-grade multi-LoRA inference serving:

{terradev_cli-5.4.1 → terradev_cli-5.4.3}/README.md

@@ -1,4 +1,4 @@
-# Terradev CLI v5.3.8
+# Terradev CLI v5.4.3
 
 **An imperative command-line-interface for AI workload orchestration.**
 
@@ -16,6 +16,29 @@ Continued focus on lower cost, faster provisioning, and topology-aware execution
 
 Model agnostic. Dataset agnostic. GPU agnostic. Provider agnostic. The only thing Terradev is not agnostic about is correctness: it enforces topology, idempotency, and sequencing.
 
+**NOTES ON 5.4.3**
+
+Fixed **credential format fallback bug** in `_provider_creds()`:
+
+- **Bug**: When credentials file contains both nested format `{"runpod": {"api_key": "..."}}` and flat format `{"runpod_api_key": "..."}`, the `_provider_creds()` method prioritized nested format even if it contained placeholder values like "PASTE_NEW_RUNPOD_KEY_HERE". This caused `terradev quote` to fail because it used placeholder credentials instead of real flat-format credentials.
+
+- **Fix**: Added placeholder detection in `_provider_creds()` to check if nested values are placeholders (containing "paste_", "your_", "example_", "test_", "placeholder_", "xxx"). If nested format contains only placeholders, the method now falls back to flat format credentials.
+
+- **Impact**: Users with mixed credential formats (nested placeholders from configure + real flat keys from onboarding) now get real credentials from the flat format, enabling `terradev quote` to return actual pricing data.
+
+**NOTES ON 5.4.2**
+
+Fixed **credential type checking bug** in `is_first_time_user()` and onboarding:
+
+- **Bug**: `is_first_time_user()` and onboarding code called `.lower()` on credential values without checking if they were strings. The credentials file uses nested dict format `{"runpod": {"api_key": "..."}}`, so when iterating over values, some are dicts, not strings. This caused `AttributeError: 'dict' object has no attribute 'lower'`.
+
+- **Fix**:
+  - Added `isinstance(value, str)` check before calling `.lower()` in `is_first_time_user()` (line 119)
+  - Added `isinstance(existing_value, str)` check before calling `.lower()` in onboarding (line 764)
+  - The `_provider_creds()` method already correctly handles both flat and nested credential formats
+
+- **Impact**: `terradev quote` and other commands that check credentials now work correctly with the nested dict credential format.
+
 **NOTES ON 5.3.9**
 
 Added **LoRAX (LoRA eXchange) integration** and **HuggingFace PEFT import** for production-grade multi-LoRA inference serving:

{terradev_cli-5.4.1 → terradev_cli-5.4.3}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "terradev-cli"
-version = "5.4.1"
+version = "5.4.3"
 description = "An imperative command-line-interface for AI workload orchestration."
 readme = "README.md"
 license = {text = "Apache-2.0"}

{terradev_cli-5.4.1 → terradev_cli-5.4.3}/terradev_cli/README.md

@@ -1,4 +1,4 @@
-# Terradev CLI v5.4.0
+# Terradev CLI v5.4.1
 
 **An imperative command-line-interface for AI workload orchestration.**
 
@@ -16,6 +16,19 @@ Continued focus on lower cost, faster provisioning, and topology-aware execution
 
 Model agnostic. Dataset agnostic. GPU agnostic. Provider agnostic. The only thing Terradev is not agnostic about is correctness: it enforces topology, idempotency, and sequencing.
 
+**NOTES ON 5.4.1**
+
+Fixed **credential persistence scope bug** in `terradev configure`:
+
+- **Bug**: `api` variable was only instantiated inside the `else` block (interactive mode), but `api.save_credentials()` was called unconditionally at line 1530. This caused `UnboundLocalError` when using `--provider` flag (single-provider mode), and would have overwritten credentials with empty data in interactive mode.
+
+- **Fix**:
+  - Moved `api = TerradevAPI()` to line 952 (before `if provider:` branching) to ensure it's always in scope
+  - Removed duplicate `api` instantiation from the `else` block
+  - Moved `api.save_credentials()` inside the `else` block (line 1531) - it only runs in interactive mode since single-provider mode saves directly to file at line 1186-1187
+
+- **Impact**: Both `terradev configure --provider <name>` and `terradev configure` (interactive) now correctly persist credentials to `~/.terradev/credentials.json`.
+
 **NOTES ON 5.4.0**
 
 Fixed **credential persistence bug** in `terradev configure` interactive mode:

{terradev_cli-5.4.1 → terradev_cli-5.4.3}/terradev_cli/cli.py

@@ -116,7 +116,7 @@ class TerradevAPI:
         # Check if all credentials are still placeholder values
         placeholder_patterns = ["your_", "example_", "test_", "placeholder_", "xxx"]
         for key, value in self.credentials.items():
-            if value and not any(
+            if value and isinstance(value, str) and not any(
                 pattern in value.lower() for pattern in placeholder_patterns
             ):
                 return False  # Found a real credential
@@ -272,8 +272,20 @@ class TerradevAPI:
         # Check for nested dict format first (from configure --provider)
         nested = self.credentials.get(provider_name, {})
         if isinstance(nested, dict) and nested:
-            # Nested dict already has the right key names for the provider
-            return {k: str(v) for k, v in nested.items() if v}
+            # Check if nested values are placeholders (not real credentials)
+            placeholder_patterns = ["paste_", "your_", "example_", "test_", "placeholder_", "xxx"]
+            has_real_creds = False
+            for v in nested.values():
+                if isinstance(v, str) and v and not any(
+                    pattern in v.lower() for pattern in placeholder_patterns
+                ):
+                    has_real_creds = True
+                    break
+            
+            if has_real_creds:
+                # Nested dict has real credentials, use it
+                return {k: str(v) for k, v in nested.items() if v}
+            # Otherwise fall back to flat format
 
         creds: Dict[str, str] = {}
         if provider_name == "aws":
@@ -761,7 +773,7 @@ def run_interactive_onboarding(api: TerradevAPI):
         existing_value = api.credentials.get(
             f"{provider_key}_api_key"
         ) or api.credentials.get(f"{provider_key}_access_key_id")
-        if existing_value and not any(
+        if existing_value and isinstance(existing_value, str) and not any(
             pattern in existing_value.lower()
             for pattern in ["your_", "example_", "test_", "placeholder_", "xxx"]
         ):

{terradev_cli-5.4.1 → terradev_cli-5.4.3}/terradev_cli.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: terradev-cli
-Version: 5.4.1
+Version: 5.4.3
 Summary: An imperative command-line-interface for AI workload orchestration.
 Author-email: Terradev Team <team@terradev.com>
 Maintainer-email: Terradev Team <team@terradev.com>
@@ -93,7 +93,7 @@ Requires-Dist: redis>=5.0.0; extra == "enterprise"
 Requires-Dist: PyJWT>=2.8.0; extra == "enterprise"
 Dynamic: license-file
 
-# Terradev CLI v5.3.8
+# Terradev CLI v5.4.3
 
 **An imperative command-line-interface for AI workload orchestration.**
 
@@ -111,6 +111,29 @@ Continued focus on lower cost, faster provisioning, and topology-aware execution
 
 Model agnostic. Dataset agnostic. GPU agnostic. Provider agnostic. The only thing Terradev is not agnostic about is correctness: it enforces topology, idempotency, and sequencing.
 
+**NOTES ON 5.4.3**
+
+Fixed **credential format fallback bug** in `_provider_creds()`:
+
+- **Bug**: When credentials file contains both nested format `{"runpod": {"api_key": "..."}}` and flat format `{"runpod_api_key": "..."}`, the `_provider_creds()` method prioritized nested format even if it contained placeholder values like "PASTE_NEW_RUNPOD_KEY_HERE". This caused `terradev quote` to fail because it used placeholder credentials instead of real flat-format credentials.
+
+- **Fix**: Added placeholder detection in `_provider_creds()` to check if nested values are placeholders (containing "paste_", "your_", "example_", "test_", "placeholder_", "xxx"). If nested format contains only placeholders, the method now falls back to flat format credentials.
+
+- **Impact**: Users with mixed credential formats (nested placeholders from configure + real flat keys from onboarding) now get real credentials from the flat format, enabling `terradev quote` to return actual pricing data.
+
+**NOTES ON 5.4.2**
+
+Fixed **credential type checking bug** in `is_first_time_user()` and onboarding:
+
+- **Bug**: `is_first_time_user()` and onboarding code called `.lower()` on credential values without checking if they were strings. The credentials file uses nested dict format `{"runpod": {"api_key": "..."}}`, so when iterating over values, some are dicts, not strings. This caused `AttributeError: 'dict' object has no attribute 'lower'`.
+
+- **Fix**:
+  - Added `isinstance(value, str)` check before calling `.lower()` in `is_first_time_user()` (line 119)
+  - Added `isinstance(existing_value, str)` check before calling `.lower()` in onboarding (line 764)
+  - The `_provider_creds()` method already correctly handles both flat and nested credential formats
+
+- **Impact**: `terradev quote` and other commands that check credentials now work correctly with the nested dict credential format.
+
 **NOTES ON 5.3.9**
 
 Added **LoRAX (LoRA eXchange) integration** and **HuggingFace PEFT import** for production-grade multi-LoRA inference serving: