terradev-cli 5.3.3__tar.gz → 5.3.4__tar.gz

{terradev_cli-5.3.3 → terradev_cli-5.3.4}/Dockerfile

@@ -18,7 +18,7 @@ RUN apt-get update && apt-get install -y \
     && rm -rf /var/lib/apt/lists/*
 
 # Copy requirements and install Python dependencies
-COPY requirements.txt .
+COPY terradev_cli/requirements.txt .
 RUN pip install --no-cache-dir -r requirements.txt
 
 # Install Terradev CLI

{terradev_cli-5.3.3 → terradev_cli-5.3.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: terradev-cli
-Version: 5.3.3
+Version: 5.3.4
 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 v5.3.3
+# Terradev CLI v5.3.4
 
 **An imperative command-line-interface for AI workload orchestration.**
 
@@ -105,9 +105,10 @@ pypi.org/project/terradev-cli/
 
 Terradev is a cross-cloud compute-provisioning CLI that compresses + stages datasets, provisions optimal instances + nodes, and deploys **3-5x faster** than sequential provisioning.
 
+
 **NOTES ON 5.3.3**
 
-Added **provider profile system** for intelligent quirk-aware routing across 23 cloud providers:
+Added **provider registration and profiling system** for intelligent quirk-aware routing across 23 cloud providers, and registration for custom providers from .yaml import:
 
 - **ProviderProfile schema** (`providers/types.py`): Encodes provider-specific behaviors including API style (REST/GraphQL/JSON:API), authentication type (Bearer/Basic/HMAC/X-Api-Key), rate limits, spot instance support, egress costs, fallback routing, capacity checks, container image pinning, and spot interruption handling.
 
@@ -126,20 +127,6 @@ Added **provider profile system** for intelligent quirk-aware routing across 23
   terradev providers export-example -o ~/.terradev/custom_providers.yaml
   ```
 
-**NOTES ON 5.3.0**
-
-Added **transport-agnostic Prefill/Decode (P/D) disaggregation layer** and **multi-agent KV cache sharing planner**:
-
-- **Transport-agnostic P/D layer** (`core/pd_transport.py`): KV cache transfer is abstracted across NIXL/NVLink (600 GB/s), NIXL/InfiniBand (200–400 GB/s), CXL 3.0 (planned migration), RoCE RDMA, and TCP fallback. The `TransportSelector` probes at provision time and selects the best available transport. Documented NIXL→CXL migration path (Phase 1 co-existence → Phase 2 CXL-primary → Phase 3 fabric switch).
-
-- **Multi-agent KV sharing planner** (`core/kv_sharing.py`): Computes fleet VRAM requirements with KV cache sharing topologies (broadcast/star/chain/none). For 20 agents with 70% shared context (broadcast), VRAM reduces by 66% → 3× more agents per GPU, 3× fewer GPUs needed, ~$14/hr savings on H100 fleet. Includes `EvictionCostModel` to quantify re-prefill penalty when VRAM is under-provisioned.
-
-- **CLI integration**: `terradev provision` now supports `--agents`, `--context`, `--sharing-topology`, `--dtype` flags. When `--agents` is passed, the KV planner outputs the heterogeneous fleet spec with cost savings:
-  ```bash
-  terradev provision -g H100 --agents 20 --context 32k --model-name llama-70b
-  terradev provision -g H100 --agents 20 --context 32k --sharing-topology broadcast --dry-run
-  ```
-
 **NOTES ON 5.2.1**
 
 Added two new BYOAPI providers: **Yotta Labs (Shakti Cloud)** and **E2E Networks** — India's leading GPU clouds. Yotta Labs uses a pod-based compute model (similar to RunPod), and E2E Networks is a traditional VM-style hyperscaler that is NSE-listed and MeitY empanelled. Both are BYOAPI: your key, stored locally, never touches a Terradev server.
@@ -771,6 +758,5 @@ Apache 2.0.
 
 ## Support
 
-- **Documentation**: [Full User Guide](USER_GUIDE.md)
 - **Issues**: [GitHub Issues](https://github.com/theoddden/Terradev/issues)
 - **Discussions**: [GitHub Discussions](https://github.com/theoddden/Terradev/discussions)

{terradev_cli-5.3.3/terradev_cli → terradev_cli-5.3.4}/README.md

@@ -1,4 +1,4 @@
-# Terradev CLI v5.3.3
+# Terradev CLI v5.3.4
 
 **An imperative command-line-interface for AI workload orchestration.**
 
@@ -10,9 +10,10 @@ pypi.org/project/terradev-cli/
 
 Terradev is a cross-cloud compute-provisioning CLI that compresses + stages datasets, provisions optimal instances + nodes, and deploys **3-5x faster** than sequential provisioning.
 
+
 **NOTES ON 5.3.3**
 
-Added **provider profile system** for intelligent quirk-aware routing across 23 cloud providers:
+Added **provider registration and profiling system** for intelligent quirk-aware routing across 23 cloud providers, and registration for custom providers from .yaml import:
 
 - **ProviderProfile schema** (`providers/types.py`): Encodes provider-specific behaviors including API style (REST/GraphQL/JSON:API), authentication type (Bearer/Basic/HMAC/X-Api-Key), rate limits, spot instance support, egress costs, fallback routing, capacity checks, container image pinning, and spot interruption handling.
 
@@ -31,20 +32,6 @@ Added **provider profile system** for intelligent quirk-aware routing across 23
   terradev providers export-example -o ~/.terradev/custom_providers.yaml
   ```
 
-**NOTES ON 5.3.0**
-
-Added **transport-agnostic Prefill/Decode (P/D) disaggregation layer** and **multi-agent KV cache sharing planner**:
-
-- **Transport-agnostic P/D layer** (`core/pd_transport.py`): KV cache transfer is abstracted across NIXL/NVLink (600 GB/s), NIXL/InfiniBand (200–400 GB/s), CXL 3.0 (planned migration), RoCE RDMA, and TCP fallback. The `TransportSelector` probes at provision time and selects the best available transport. Documented NIXL→CXL migration path (Phase 1 co-existence → Phase 2 CXL-primary → Phase 3 fabric switch).
-
-- **Multi-agent KV sharing planner** (`core/kv_sharing.py`): Computes fleet VRAM requirements with KV cache sharing topologies (broadcast/star/chain/none). For 20 agents with 70% shared context (broadcast), VRAM reduces by 66% → 3× more agents per GPU, 3× fewer GPUs needed, ~$14/hr savings on H100 fleet. Includes `EvictionCostModel` to quantify re-prefill penalty when VRAM is under-provisioned.
-
-- **CLI integration**: `terradev provision` now supports `--agents`, `--context`, `--sharing-topology`, `--dtype` flags. When `--agents` is passed, the KV planner outputs the heterogeneous fleet spec with cost savings:
-  ```bash
-  terradev provision -g H100 --agents 20 --context 32k --model-name llama-70b
-  terradev provision -g H100 --agents 20 --context 32k --sharing-topology broadcast --dry-run
-  ```
-
 **NOTES ON 5.2.1**
 
 Added two new BYOAPI providers: **Yotta Labs (Shakti Cloud)** and **E2E Networks** — India's leading GPU clouds. Yotta Labs uses a pod-based compute model (similar to RunPod), and E2E Networks is a traditional VM-style hyperscaler that is NSE-listed and MeitY empanelled. Both are BYOAPI: your key, stored locally, never touches a Terradev server.
@@ -676,6 +663,5 @@ Apache 2.0.
 
 ## Support
 
-- **Documentation**: [Full User Guide](USER_GUIDE.md)
 - **Issues**: [GitHub Issues](https://github.com/theoddden/Terradev/issues)
 - **Discussions**: [GitHub Discussions](https://github.com/theoddden/Terradev/discussions)

{terradev_cli-5.3.3 → terradev_cli-5.3.4}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "terradev-cli"
-version = "5.3.3"
+version = "5.3.4"
 description = "An imperative command-line-interface for AI workload orchestration."
 readme = "README.md"
 license = {text = "Apache-2.0"}

{terradev_cli-5.3.3 → terradev_cli-5.3.4}/terradev_cli/__init__.py

@@ -3,6 +3,6 @@ Terradev CLI - Cross-Cloud Compute Optimization Platform
 Parallel provisioning and orchestration for optimized compute costs
 """
 
-__version__ = "5.0.1"
+__version__ = "5.3.4"
 __author__ = "Terradev Team"
 __description__ = "Open Source BYOAPI Multi-Cloud GPU Infrastructure Platform"

{terradev_cli-5.3.3 → terradev_cli-5.3.4}/terradev_cli/cli.py

@@ -14138,11 +14138,11 @@ def local_scan(host, user, key, detailed, register, name):
             except Exception:
                 return "", 1
         else:
-            # Local execution - still safe since query is hardcoded
-            cmd = f"nvidia-smi --query-gpu={query} --format=csv,noheader,nounits"
+            # Local execution - use list-args for injection safety
+            cmd = ["nvidia-smi", f"--query-gpu={query}", "--format=csv,noheader,nounits"]
             try:
                 result = subprocess.run(
-                    cmd, shell=True, capture_output=True, text=True, timeout=15
+                    cmd, capture_output=True, text=True, timeout=15
                 )
                 return result.stdout.strip(), result.returncode
             except Exception:
@@ -14318,11 +14318,11 @@ def local_register(name, host, user, key):
             click.echo(f"Error scanning {target}: {e}", err=True)
             return
     else:
-        # Local execution - safe since query is hardcoded
-        cmd = f"nvidia-smi --query-gpu={query} --format=csv,noheader,nounits"
+        # Local execution - use list-args for injection safety
+        cmd = ["nvidia-smi", f"--query-gpu={query}", "--format=csv,noheader,nounits"]
         try:
             result = subprocess.run(
-                cmd, shell=True, capture_output=True, text=True, timeout=15
+                cmd, capture_output=True, text=True, timeout=15
             )
             raw = result.stdout.strip()
         except Exception as e:

{terradev_cli-5.3.3 → terradev_cli-5.3.4}/terradev_cli/core/checkpoint_manager.py

@@ -221,7 +221,8 @@ def _run_on(
     import re
 
     # Allow alphanumeric, spaces, and basic shell-safe characters
-    if not re.match(r'^[a-zA-Z0-9_\-./:=@, \n\t|>&]+$', cmd):
+    # Explicitly exclude shell metacharacters (|, >, &) to prevent command chaining
+    if not re.match(r'^[a-zA-Z0-9_\-./:=@, \n\t]+$', cmd):
         return -1, "", "Unsafe command characters detected"
 
     if host and host not in ("localhost", "127.0.0.1"):

{terradev_cli-5.3.3 → terradev_cli-5.3.4}/terradev_cli/core/preflight_validator.py

@@ -139,7 +139,8 @@ def _run_on(
     import re
 
     # Allow alphanumeric, spaces, and basic shell-safe characters
-    if not re.match(r'^[a-zA-Z0-9_\-./:=@, \n\t|>&]+$', cmd):
+    # Explicitly exclude shell metacharacters (|, >, &) to prevent command chaining
+    if not re.match(r'^[a-zA-Z0-9_\-./:=@, \n\t]+$', cmd):
         return -1, "", "Unsafe command characters detected"
 
     if host and host not in ("localhost", "127.0.0.1"):

{terradev_cli-5.3.3 → terradev_cli-5.3.4}/terradev_cli/core/training_monitor.py

@@ -136,7 +136,8 @@ def _run_on(
     import re
 
     # Allow alphanumeric, spaces, and basic shell-safe characters
-    if not re.match(r'^[a-zA-Z0-9_\-./:=@, \n\t|>&]+$', cmd):
+    # Explicitly exclude shell metacharacters (|, >, &) to prevent command chaining
+    if not re.match(r'^[a-zA-Z0-9_\-./:=@, \n\t]+$', cmd):
         return -1, "", "Unsafe command characters detected"
 
     if host and host not in ("localhost", "127.0.0.1"):

{terradev_cli-5.3.3 → terradev_cli-5.3.4}/terradev_cli/core/training_orchestrator.py

@@ -439,7 +439,8 @@ def _run_on(
     # Allow alphanumeric, spaces, and basic shell-safe characters.
     # SECURITY: quotes ('"') are deliberately excluded — they are shell metacharacters
     # that can be used to escape a quoted context and inject arbitrary commands.
-    if not re.match(r'^[a-zA-Z0-9_\-./:=@, \n\t|>&]+$', cmd):
+    # Explicitly exclude shell metacharacters (|, >, &) to prevent command chaining
+    if not re.match(r'^[a-zA-Z0-9_\-./:=@, \n\t]+$', cmd):
         return -1, "", "Unsafe command characters detected"
 
     if host and host not in ("localhost", "127.0.0.1"):

{terradev_cli-5.3.3 → terradev_cli-5.3.4}/terradev_cli/mcp/server.py

@@ -2,7 +2,7 @@
 """
 Terradev MCP Server v2.1.0 - Complete Agentic GPU Infrastructure for Claude Code
 
-218 MCP tools: GPU provisioning, Kubernetes clusters, Karpenter auto-provisioning,
+220 MCP tools: GPU provisioning, Kubernetes clusters, Karpenter auto-provisioning,
 GitOps/ArgoCD automation, event-driven triggers, environment promotion, lineage tracking,
 cross-provider migration, vLLM/SGLang/Ollama inference, Arize Phoenix trace observability,
 NeMo Guardrails output safety, Qdrant vector DB, Ray cluster management, W&B/LangSmith/MLflow/DVC,
@@ -982,15 +982,19 @@ server = Server("terradev-mcp")
 _tools_loaded = False
 _tool_registry: Dict[str, callable] = {}
 _tool_schemas: Dict[str, Dict] = {}
-_load_lock = asyncio.Lock()
+_load_lock: Optional[asyncio.Lock] = None
 _concurrent_requests = 0
 _max_concurrent = 100
-_request_semaphore = asyncio.Semaphore(_max_concurrent)
+_request_semaphore: Optional[asyncio.Semaphore] = None
 
 
 async def _ensure_tools_loaded():
     """Ensure tools are loaded (lazy loading)"""
-    global _tools_loaded, _tool_registry, _tool_schemas
+    global _tools_loaded, _tool_registry, _tool_schemas, _load_lock
+
+    # Lazy-create lock to avoid Python 3.9 event loop binding bug
+    if _load_lock is None:
+        _load_lock = asyncio.Lock()
 
     if _tools_loaded:
         return
@@ -5336,7 +5340,11 @@ else:
 @server.list_tools()
 async def handle_list_tools() -> ListToolsResult:
     """List available Terradev tools (lazy loading with compression)"""
-    global _concurrent_requests
+    global _concurrent_requests, _request_semaphore
+
+    # Lazy-create semaphore to avoid Python 3.9 event loop binding bug
+    if _request_semaphore is None:
+        _request_semaphore = asyncio.Semaphore(_max_concurrent)
 
     # Adaptive concurrency control
     async with _request_semaphore:
@@ -5355,11 +5363,15 @@ async def handle_list_tools() -> ListToolsResult:
 @server.call_tool()
 async def handle_call_tool(request: CallToolRequest) -> CallToolResult:
     """Handle tool calls with adaptive concurrency control"""
-    global _concurrent_requests
+    global _concurrent_requests, _request_semaphore
 
     tool_name = request.params.name
     arguments = request.params.arguments or {}
 
+    # Lazy-create semaphore to avoid Python 3.9 event loop binding bug
+    if _request_semaphore is None:
+        _request_semaphore = asyncio.Semaphore(_max_concurrent)
+
     # Adaptive concurrency control
     async with _request_semaphore:
         _concurrent_requests += 1

{terradev_cli-5.3.3 → terradev_cli-5.3.4}/terradev_cli/providers/__init__.py

@@ -18,6 +18,7 @@ from .types import (
     ProviderCapabilities,
     HealthStatus,
     ProviderHealth,
+    ProviderProfile,
 )
 
 from .base_provider import BaseProvider

{terradev_cli-5.3.3 → terradev_cli-5.3.4}/terradev_cli/providers/base_provider.py

@@ -42,11 +42,15 @@ class BaseProvider(ABC):
 
     # Shared TCP connector for connection pooling across all providers
     _shared_connector: Optional[aiohttp.TCPConnector] = None
-    _connector_lock = asyncio.Lock()
+    _connector_lock: Optional[asyncio.Lock] = None
 
     @classmethod
     async def _get_shared_connector(cls) -> aiohttp.TCPConnector:
         """Get or create shared TCP connector with connection pooling"""
+        # Lazy-create lock to avoid Python 3.9 event loop binding bug
+        if cls._connector_lock is None:
+            cls._connector_lock = asyncio.Lock()
+
         if cls._shared_connector is None or cls._shared_connector.closed:
             async with cls._connector_lock:
                 if cls._shared_connector is None or cls._shared_connector.closed:
@@ -370,7 +374,7 @@ class BaseProvider(ABC):
                         last_states[iid] = info.status
                     except Exception as e:
                         # Log but don't break the loop
-                        pass
+                        logger.debug(f"poll_loop error for {iid}: {e}")
 
                 await asyncio.sleep(poll_interval_s)
 

{terradev_cli-5.3.3 → terradev_cli-5.3.4}/terradev_cli/providers/e2e_networks_provider.py

@@ -115,7 +115,7 @@ class E2ENetworksProvider(BaseProvider):
                 "instance_type": self.GPU_PLAN_MAP.get(gpu_type, gpu_type.lower()),
                 "gpu_type": gpu_type,
                 "price_per_hour": info["price"],
-                "region": region or "in-delhi-1",
+                "region": region or "in-mumbai-1",
                 "available": True,
                 "provider": "e2enetworks",
                 "vcpus": info["vcpus"],
@@ -156,7 +156,7 @@ class E2ENetworksProvider(BaseProvider):
             # Convert INR to USD if price looks like INR (>10 per hour for GPU)
             price_usd = price_inr / 83.5 if price_inr > 10 else price_inr
 
-            plan_region = plan.get("region", region or "in-delhi-1")
+            plan_region = plan.get("region", region or "in-mumbai-1")
             if region and plan_region != region:
                 continue
 
@@ -195,7 +195,7 @@ class E2ENetworksProvider(BaseProvider):
         body: Dict[str, Any] = {
             "name": f"terradev-{gpu_type.lower().replace('_', '-')}-{datetime.now().strftime('%H%M%S')}",
             "plan": instance_type,
-            "location": region or "in-delhi-1",
+            "location": region or "in-mumbai-1",
             "image": image,
             "gpu_count": gpu_count,
         }
@@ -208,7 +208,7 @@ class E2ENetworksProvider(BaseProvider):
         return {
             "instance_id": str(node.get("id", f"e2e-{datetime.now().strftime('%Y%m%d%H%M%S')}")),
             "instance_type": instance_type,
-            "region": region or "in-delhi-1",
+            "region": region or "in-mumbai-1",
             "gpu_type": gpu_type,
             "status": node.get("status", "provisioning"),
             "provider": "e2enetworks",

{terradev_cli-5.3.3 → terradev_cli-5.3.4}/terradev_cli/providers/fluidstack_provider.py

@@ -135,7 +135,7 @@ class FluidStackProvider(BaseProvider):
                 "instance_type": info["gpu_type"],
                 "gpu_type": gpu_type,
                 "price_per_hour": info["price"],
-                "region": region or "us-east-1",
+                "region": region or "uk-london",
                 "available": True,
                 "provider": "fluidstack",
                 "vcpus": info["vcpus"],
@@ -173,7 +173,7 @@ class FluidStackProvider(BaseProvider):
                         "instance_type": config_gpu,
                         "gpu_type": gpu_type,
                         "price_per_hour": price,
-                        "region": plan.get("region", "us-east-1"),
+                        "region": plan.get("region", "uk-london"),
                         "available": plan.get("available", True),
                         "provider": "fluidstack",
                         "vcpus": config.get("vcpus", plan.get("vcpus", 16)),

{terradev_cli-5.3.3 → terradev_cli-5.3.4}/terradev_cli/providers/registry.py

@@ -15,6 +15,7 @@ Benefits:
 
 import time
 import logging
+import asyncio
 from typing import Dict, List, Optional
 from collections import defaultdict
 
@@ -48,7 +49,7 @@ class ProviderRegistry:
             lambda: ProviderHealth(provider="")
         )
         self._spot_preemptions: Dict[str, List[float]] = defaultdict(list)  # provider → timestamps
-        self._lock = None  # TODO: add threading lock if needed
+        self._lock: Optional[asyncio.Lock] = None
 
     def is_healthy(self, provider: str) -> bool:
         """
@@ -71,57 +72,72 @@ class ProviderRegistry:
 
         return True
 
-    def record_success(self, provider: str, latency_ms: float):
+    async def record_success(self, provider: str, latency_ms: float):
         """
         Record a successful operation for a provider.
 
         Resets consecutive failure counter and updates latency metrics.
         """
-        health = self._health[provider]
-        health.provider = provider
-        health.consecutive_failures = 0
-        health.last_success_ts = time.time()
-        health.total_provisions += 1
+        # Lazy-create lock to avoid Python 3.9 event loop binding bug
+        if self._lock is None:
+            self._lock = asyncio.Lock()
 
-        # Exponential moving average for latency
-        if health.avg_latency_ms == 0:
-            health.avg_latency_ms = latency_ms
-        else:
-            health.avg_latency_ms = 0.9 * health.avg_latency_ms + 0.1 * latency_ms
+        async with self._lock:
+            health = self._health[provider]
+            health.provider = provider
+            health.consecutive_failures = 0
+            health.last_success_ts = time.time()
+            health.total_provisions += 1
 
-    def record_failure(self, provider: str, error: str = ""):
+            # Exponential moving average for latency
+            if health.avg_latency_ms == 0:
+                health.avg_latency_ms = latency_ms
+            else:
+                health.avg_latency_ms = 0.9 * health.avg_latency_ms + 0.1 * latency_ms
+
+    async def record_failure(self, provider: str, error: str = ""):
         """
         Record a failed operation for a provider.
 
         Increments consecutive failure counter. Opens circuit breaker
         after FAILURE_THRESHOLD consecutive failures.
         """
-        health = self._health[provider]
-        health.provider = provider
-        health.consecutive_failures += 1
-        health.last_failure_ts = time.time()
-        health.total_failures += 1
+        # Lazy-create lock to avoid Python 3.9 event loop binding bug
+        if self._lock is None:
+            self._lock = asyncio.Lock()
 
-        if health.consecutive_failures >= self.FAILURE_THRESHOLD:
-            logger.warning(
-                f"Circuit breaker opened for provider: {provider} "
-                f"(failures: {health.consecutive_failures}, error: {error})"
-            )
+        async with self._lock:
+            health = self._health[provider]
+            health.provider = provider
+            health.consecutive_failures += 1
+            health.last_failure_ts = time.time()
+            health.total_failures += 1
 
-    def record_preemption(self, provider: str, region: str = ""):
+            if health.consecutive_failures >= self.FAILURE_THRESHOLD:
+                logger.warning(
+                    f"Circuit breaker opened for provider: {provider} "
+                    f"(failures: {health.consecutive_failures}, error: {error})"
+                )
+
+    async def record_preemption(self, provider: str, region: str = ""):
         """
         Record a spot instance preemption for a provider.
 
         Used to calculate spot preemption rate for provider ranking.
         """
-        now = time.time()
-        self._spot_preemptions[provider].append(now)
-
-        # Clean old preemptions (older than 24 hours)
-        cutoff = now - 86400
-        self._spot_preemptions[provider] = [
-            ts for ts in self._spot_preemptions[provider] if ts > cutoff
-        ]
+        # Lazy-create lock to avoid Python 3.9 event loop binding bug
+        if self._lock is None:
+            self._lock = asyncio.Lock()
+
+        async with self._lock:
+            now = time.time()
+            self._spot_preemptions[provider].append(now)
+
+            # Clean old preemptions (older than 24 hours)
+            cutoff = now - 86400
+            self._spot_preemptions[provider] = [
+                ts for ts in self._spot_preemptions[provider] if ts > cutoff
+            ]
 
     def get_spot_score(self, provider: str, region: str = "") -> float:
         """

{terradev_cli-5.3.3 → terradev_cli-5.3.4}/terradev_cli/providers/yottalabs_provider.py

@@ -142,7 +142,7 @@ class YottaLabsProvider(BaseProvider):
                 "instance_type": yotta_gpu_type,
                 "gpu_type": gpu_type,
                 "price_per_hour": info["price"],
-                "region": region or "us-east-1",
+                "region": region or "ap-south-1",
                 "available": True,
                 "provider": "yottalabs",
                 "vcpus": info["vcpus"],
@@ -170,8 +170,8 @@ class YottaLabsProvider(BaseProvider):
                 ):
                     continue
 
-            for r in entry.get("regions", [region or "us-east-1"]):
-                target_region = r if isinstance(r, str) else r.get("region", "us-east-1")
+            for r in entry.get("regions", [region or "ap-south-1"]):
+                target_region = r if isinstance(r, str) else r.get("region", "ap-south-1")
                 if region and target_region != region:
                     continue
                 price = entry.get("pricePerHour", entry.get("price_per_hour", 0))
@@ -213,7 +213,7 @@ class YottaLabsProvider(BaseProvider):
             "gpuType": yotta_gpu_type,
             "gpuCount": gpu_count,
             "containerVolumeInGb": volume_gb,
-            "regions": [region] if region else ["us-east-1"],
+            "regions": [region] if region else ["ap-south-1"],
             "minSingleCardVcpu": 8,
             "minSingleCardRamInGb": 32,
             "environmentVars": [
@@ -229,7 +229,7 @@ class YottaLabsProvider(BaseProvider):
         return {
             "instance_id": str(pod.get("id", f"yl-{datetime.now().strftime('%Y%m%d%H%M%S')}")),
             "instance_type": yotta_gpu_type,
-            "region": region or "us-east-1",
+            "region": region or "ap-south-1",
             "gpu_type": gpu_type,
             "status": pod.get("status", "Initialize"),
             "provider": "yottalabs",

{terradev_cli-5.3.3 → terradev_cli-5.3.4}/terradev_cli.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: terradev-cli
-Version: 5.3.3
+Version: 5.3.4
 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 v5.3.3
+# Terradev CLI v5.3.4
 
 **An imperative command-line-interface for AI workload orchestration.**
 
@@ -105,9 +105,10 @@ pypi.org/project/terradev-cli/
 
 Terradev is a cross-cloud compute-provisioning CLI that compresses + stages datasets, provisions optimal instances + nodes, and deploys **3-5x faster** than sequential provisioning.
 
+
 **NOTES ON 5.3.3**
 
-Added **provider profile system** for intelligent quirk-aware routing across 23 cloud providers:
+Added **provider registration and profiling system** for intelligent quirk-aware routing across 23 cloud providers, and registration for custom providers from .yaml import:
 
 - **ProviderProfile schema** (`providers/types.py`): Encodes provider-specific behaviors including API style (REST/GraphQL/JSON:API), authentication type (Bearer/Basic/HMAC/X-Api-Key), rate limits, spot instance support, egress costs, fallback routing, capacity checks, container image pinning, and spot interruption handling.
 
@@ -126,20 +127,6 @@ Added **provider profile system** for intelligent quirk-aware routing across 23
   terradev providers export-example -o ~/.terradev/custom_providers.yaml
   ```
 
-**NOTES ON 5.3.0**
-
-Added **transport-agnostic Prefill/Decode (P/D) disaggregation layer** and **multi-agent KV cache sharing planner**:
-
-- **Transport-agnostic P/D layer** (`core/pd_transport.py`): KV cache transfer is abstracted across NIXL/NVLink (600 GB/s), NIXL/InfiniBand (200–400 GB/s), CXL 3.0 (planned migration), RoCE RDMA, and TCP fallback. The `TransportSelector` probes at provision time and selects the best available transport. Documented NIXL→CXL migration path (Phase 1 co-existence → Phase 2 CXL-primary → Phase 3 fabric switch).
-
-- **Multi-agent KV sharing planner** (`core/kv_sharing.py`): Computes fleet VRAM requirements with KV cache sharing topologies (broadcast/star/chain/none). For 20 agents with 70% shared context (broadcast), VRAM reduces by 66% → 3× more agents per GPU, 3× fewer GPUs needed, ~$14/hr savings on H100 fleet. Includes `EvictionCostModel` to quantify re-prefill penalty when VRAM is under-provisioned.
-
-- **CLI integration**: `terradev provision` now supports `--agents`, `--context`, `--sharing-topology`, `--dtype` flags. When `--agents` is passed, the KV planner outputs the heterogeneous fleet spec with cost savings:
-  ```bash
-  terradev provision -g H100 --agents 20 --context 32k --model-name llama-70b
-  terradev provision -g H100 --agents 20 --context 32k --sharing-topology broadcast --dry-run
-  ```
-
 **NOTES ON 5.2.1**
 
 Added two new BYOAPI providers: **Yotta Labs (Shakti Cloud)** and **E2E Networks** — India's leading GPU clouds. Yotta Labs uses a pod-based compute model (similar to RunPod), and E2E Networks is a traditional VM-style hyperscaler that is NSE-listed and MeitY empanelled. Both are BYOAPI: your key, stored locally, never touches a Terradev server.
@@ -771,6 +758,5 @@ Apache 2.0.
 
 ## Support
 
-- **Documentation**: [Full User Guide](USER_GUIDE.md)
 - **Issues**: [GitHub Issues](https://github.com/theoddden/Terradev/issues)
 - **Discussions**: [GitHub Discussions](https://github.com/theoddden/Terradev/discussions)

{terradev_cli-5.3.3 → terradev_cli-5.3.4}/terradev_cli.egg-info/SOURCES.txt

@@ -5,7 +5,6 @@ LICENSE
 MANIFEST.in
 README.md
 pyproject.toml
-requirements.txt
 terraform.tfvars.example
 .github/workflows/action.yml
 .github/workflows/ci.yml

terradev_cli-5.3.3/requirements.txt

@@ -1,56 +0,0 @@
-# Terradev Python Dependencies
-
-# Cloud SDKs
-boto3==1.28.57
-botocore==1.31.57
-google-cloud-billing==1.10.0
-azure-mgmt-billing==6.0.0
-aiohttp==3.9.1
-
-# Data Processing
-pandas==2.0.3
-numpy==1.24.3
-requests==2.31.0
-
-# Machine Learning
-tensorflow==2.13.0
-torch==2.0.1
-scikit-learn==1.3.0
-keras==2.13.1
-transformers==4.35.2
-datasets==2.14.6
-tokenizers==0.14.1
-
-# Hugging Face Integration
-huggingface-hub==0.17.3
-
-# Data Visualization
-matplotlib==3.7.2
-seaborn==0.12.2
-
-# Data Feed Processing
-feedparser==6.0.10
-beautifulsoup4==4.12.2
-lxml==4.9.3
-
-# Monitoring and Logging
-psutil==5.9.5
-gpustat==1.1.1
-prometheus-client==0.17.1
-
-# Utilities
-click==8.1.7
-pyyaml==6.0.1
-jinja2==3.1.2
-python-dotenv==1.0.0
-
-# Development and Testing
-pytest==7.4.2
-pytest-cov==4.1.0
-black==23.7.0
-flake8==6.0.0
-mypy==1.5.1
-
-# Documentation
-sphinx==7.1.2
-sphinx-rtd-theme==1.3.0