Sync with upstream ollama/ollama and restore Tesla K80 (compute 3.7) support

This commit represents a complete rework after pulling the latest changes from
official ollama/ollama repository and re-applying Tesla K80 compatibility patches.

## Key Changes

### CUDA Compute Capability 3.7 Support (Tesla K80)
- Added sm_37 (compute 3.7) to CMAKE_CUDA_ARCHITECTURES in CMakeLists.txt
- Updated CMakePresets.json to include compute 3.7 in "CUDA 11" preset
- Using 37-virtual (PTX with JIT compilation) for maximum compatibility

### Legacy Toolchain Compatibility
- **NVIDIA Driver**: 470.256.02 (last version supporting Kepler/K80)
- **CUDA Version**: 11.4.4 (last CUDA 11.x supporting compute 3.7)
- **GCC Version**: 10.5.0 (required by CUDA 11.4 host_config.h)

### CPU Architecture Trade-offs
Due to GCC 10.5 limitation, sacrificed newer CPU optimizations:
- Alderlake CPU variant enabled WITHOUT AVX_VNNI (requires GCC 11+)
- Still supports: SSE4.2, AVX, F16C, AVX2, BMI2, FMA
- Performance impact: ~3-7% on newer CPUs (acceptable for K80 compatibility)

### Build System Updates
- Modified ml/backend/ggml/ggml/src/ggml-cuda/CMakeLists.txt for compute 3.7
- Added -Wno-deprecated-gpu-targets flag to suppress warnings
- Updated ml/backend/ggml/ggml/src/CMakeLists.txt for Alderlake without AVX_VNNI

### Upstream Sync
Merged latest llama.cpp changes including:
- Enhanced KV cache management with ISWA and hybrid memory support
- Improved multi-modal support (mtmd framework)
- New model architectures (Gemma3, Llama4, Qwen3, etc.)
- GPU backend improvements for CUDA, Metal, and ROCm
- Updated quantization support and GGUF format handling

### Documentation
- Updated CLAUDE.md with comprehensive build instructions
- Documented toolchain constraints and CPU architecture trade-offs
- Removed outdated CI/CD workflows (tesla-k80-*.yml)
- Cleaned up temporary development artifacts

## Rationale

This fork maintains Tesla K80 GPU support (compute 3.7) which was dropped in
official Ollama due to legacy driver/CUDA requirements. The toolchain constraint
creates a deadlock:
- K80 → Driver 470 → CUDA 11.4 → GCC 10 → No AVX_VNNI

We accept the loss of cutting-edge CPU optimizations to enable running modern
LLMs on legacy but still capable Tesla K80 hardware (12GB VRAM per GPU).

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

api/types.go

@@ -11,6 +11,8 @@ import (
 	"strings"
 	"time"
 
+	"github.com/google/uuid"
+
 	"github.com/ollama/ollama/envconfig"
 	"github.com/ollama/ollama/types/model"
 )
@@ -36,6 +38,19 @@ func (e StatusError) Error() string {
 	}
 }
 
+type AuthorizationError struct {
+	StatusCode int
+	Status     string
+	SigninURL  string `json:"signin_url"`
+}
+
+func (e AuthorizationError) Error() string {
+	if e.Status != "" {
+		return e.Status
+	}
+	return "something went wrong, please see the ollama server logs for details"
+}
+
 // ImageData represents the raw binary data of an image file.
 type ImageData []byte
 
@@ -90,6 +105,18 @@ type GenerateRequest struct {
 	// (request that thinking _not_ be used) and unset (use the old behavior
 	// before this option was introduced)
 	Think *ThinkValue `json:"think,omitempty"`
+
+	// Truncate is a boolean that, when set to true, truncates the chat history messages
+	// if the rendered prompt exceeds the context length limit.
+	Truncate *bool `json:"truncate,omitempty"`
+
+	// Shift is a boolean that, when set to true, shifts the chat history
+	// when hitting the context length limit instead of erroring.
+	Shift *bool `json:"shift,omitempty"`
+
+	// DebugRenderOnly is a debug option that, when set to true, returns the rendered
+	// template instead of calling the model.
+	DebugRenderOnly bool `json:"_debug_render_only,omitempty"`
 }
 
 // ChatRequest describes a request sent by [Client.Chat].
@@ -120,6 +147,18 @@ type ChatRequest struct {
 	// responding. Can be a boolean (true/false) or a string ("high", "medium", "low")
 	// for supported models.
 	Think *ThinkValue `json:"think,omitempty"`
+
+	// Truncate is a boolean that, when set to true, truncates the chat history messages
+	// if the rendered prompt exceeds the context length limit.
+	Truncate *bool `json:"truncate,omitempty"`
+
+	// Shift is a boolean that, when set to true, shifts the chat history
+	// when hitting the context length limit instead of erroring.
+	Shift *bool `json:"shift,omitempty"`
+
+	// DebugRenderOnly is a debug option that, when set to true, returns the rendered
+	// template instead of calling the model.
+	DebugRenderOnly bool `json:"_debug_render_only,omitempty"`
 }
 
 type Tools []Tool
@@ -165,7 +204,7 @@ type ToolCall struct {
 }
 
 type ToolCallFunction struct {
-	Index     int                       `json:"index,omitempty"`
+	Index     int                       `json:"index"`
 	Name      string                    `json:"name"`
 	Arguments ToolCallFunctionArguments `json:"arguments"`
 }
@@ -227,9 +266,9 @@ func (pt PropertyType) String() string {
 
 type ToolProperty struct {
 	AnyOf       []ToolProperty `json:"anyOf,omitempty"`
-	Type        PropertyType   `json:"type"`
+	Type        PropertyType   `json:"type,omitempty"`
 	Items       any            `json:"items,omitempty"`
-	Description string         `json:"description"`
+	Description string         `json:"description,omitempty"`
 	Enum        []any          `json:"enum,omitempty"`
 }
 
@@ -278,16 +317,23 @@ func mapToTypeScriptType(jsonType string) string {
 	}
 }
 
+type ToolFunctionParameters struct {
+	Type       string                  `json:"type"`
+	Defs       any                     `json:"$defs,omitempty"`
+	Items      any                     `json:"items,omitempty"`
+	Required   []string                `json:"required"`
+	Properties map[string]ToolProperty `json:"properties"`
+}
+
+func (t *ToolFunctionParameters) String() string {
+	bts, _ := json.Marshal(t)
+	return string(bts)
+}
+
 type ToolFunction struct {
-	Name        string `json:"name"`
-	Description string `json:"description"`
-	Parameters  struct {
-		Type       string                  `json:"type"`
-		Defs       any                     `json:"$defs,omitempty"`
-		Items      any                     `json:"items,omitempty"`
-		Required   []string                `json:"required"`
-		Properties map[string]ToolProperty `json:"properties"`
-	} `json:"parameters"`
+	Name        string                 `json:"name"`
+	Description string                 `json:"description,omitempty"`
+	Parameters  ToolFunctionParameters `json:"parameters"`
 }
 
 func (t *ToolFunction) String() string {
@@ -298,16 +344,38 @@ func (t *ToolFunction) String() string {
 // ChatResponse is the response returned by [Client.Chat]. Its fields are
 // similar to [GenerateResponse].
 type ChatResponse struct {
-	Model      string    `json:"model"`
-	CreatedAt  time.Time `json:"created_at"`
-	Message    Message   `json:"message"`
-	DoneReason string    `json:"done_reason,omitempty"`
+	// Model is the model name that generated the response.
+	Model string `json:"model"`
 
+	// RemoteModel is the name of the upstream model that generated the response.
+	RemoteModel string `json:"remote_model,omitempty"`
+
+	// RemoteHost is the URL of the upstream Ollama host that generated the response.
+	RemoteHost string `json:"remote_host,omitempty"`
+
+	// CreatedAt is the timestamp of the response.
+	CreatedAt time.Time `json:"created_at"`
+
+	// Message contains the message or part of a message from the model.
+	Message Message `json:"message"`
+
+	// Done specifies if the response is complete.
 	Done bool `json:"done"`
 
+	// DoneReason is the reason the model stopped generating text.
+	DoneReason string `json:"done_reason,omitempty"`
+
+	DebugInfo *DebugInfo `json:"_debug_info,omitempty"`
+
 	Metrics
 }
 
+// DebugInfo contains debug information for template rendering
+type DebugInfo struct {
+	RenderedTemplate string `json:"rendered_template"`
+	ImageCount       int    `json:"image_count,omitempty"`
+}
+
 type Metrics struct {
 	TotalDuration      time.Duration `json:"total_duration,omitempty"`
 	LoadDuration       time.Duration `json:"load_duration,omitempty"`
@@ -360,8 +428,12 @@ type EmbedRequest struct {
 	// this request.
 	KeepAlive *Duration `json:"keep_alive,omitempty"`
 
+	// Truncate truncates the input to fit the model's max sequence length.
 	Truncate *bool `json:"truncate,omitempty"`
 
+	// Dimensions truncates the output embedding to the specified dimension.
+	Dimensions int `json:"dimensions,omitempty"`
+
 	// Options lists model-specific options.
 	Options map[string]any `json:"options"`
 }
@@ -399,18 +471,47 @@ type EmbeddingResponse struct {
 
 // CreateRequest is the request passed to [Client.Create].
 type CreateRequest struct {
-	Model    string `json:"model"`
-	Stream   *bool  `json:"stream,omitempty"`
+	// Model is the model name to create.
+	Model string `json:"model"`
+
+	// Stream specifies whether the response is streaming; it is true by default.
+	Stream *bool `json:"stream,omitempty"`
+
+	// Quantize is the quantization format for the model; leave blank to not change the quantization level.
 	Quantize string `json:"quantize,omitempty"`
 
-	From       string            `json:"from,omitempty"`
-	Files      map[string]string `json:"files,omitempty"`
-	Adapters   map[string]string `json:"adapters,omitempty"`
-	Template   string            `json:"template,omitempty"`
-	License    any               `json:"license,omitempty"`
-	System     string            `json:"system,omitempty"`
-	Parameters map[string]any    `json:"parameters,omitempty"`
-	Messages   []Message         `json:"messages,omitempty"`
+	// From is the name of the model or file to use as the source.
+	From string `json:"from,omitempty"`
+
+	// RemoteHost is the URL of the upstream ollama API for the model (if any).
+	RemoteHost string `json:"remote_host,omitempty"`
+
+	// Files is a map of files include when creating the model.
+	Files map[string]string `json:"files,omitempty"`
+
+	// Adapters is a map of LoRA adapters to include when creating the model.
+	Adapters map[string]string `json:"adapters,omitempty"`
+
+	// Template is the template used when constructing a request to the model.
+	Template string `json:"template,omitempty"`
+
+	// License is a string or list of strings for licenses.
+	License any `json:"license,omitempty"`
+
+	// System is the system prompt for the model.
+	System string `json:"system,omitempty"`
+
+	// Parameters is a map of hyper-parameters which are applied to the model.
+	Parameters map[string]any `json:"parameters,omitempty"`
+
+	// Messages is a list of messages added to the model before chat and generation requests.
+	Messages []Message `json:"messages,omitempty"`
+
+	Renderer string `json:"renderer,omitempty"`
+	Parser   string `json:"parser,omitempty"`
+
+	// Info is a map of additional information for the model
+	Info map[string]any `json:"info,omitempty"`
 
 	// Deprecated: set the model name with Model instead
 	Name string `json:"name"`
@@ -448,8 +549,12 @@ type ShowResponse struct {
 	Parameters    string             `json:"parameters,omitempty"`
 	Template      string             `json:"template,omitempty"`
 	System        string             `json:"system,omitempty"`
+	Renderer      string             `json:"renderer,omitempty"`
+	Parser        string             `json:"parser,omitempty"`
 	Details       ModelDetails       `json:"details,omitempty"`
 	Messages      []Message          `json:"messages,omitempty"`
+	RemoteModel   string             `json:"remote_model,omitempty"`
+	RemoteHost    string             `json:"remote_host,omitempty"`
 	ModelInfo     map[string]any     `json:"model_info,omitempty"`
 	ProjectorInfo map[string]any     `json:"projector_info,omitempty"`
 	Tensors       []Tensor           `json:"tensors,omitempty"`
@@ -508,12 +613,14 @@ type ProcessResponse struct {
 
 // ListModelResponse is a single model description in [ListResponse].
 type ListModelResponse struct {
-	Name       string       `json:"name"`
-	Model      string       `json:"model"`
-	ModifiedAt time.Time    `json:"modified_at"`
-	Size       int64        `json:"size"`
-	Digest     string       `json:"digest"`
-	Details    ModelDetails `json:"details,omitempty"`
+	Name        string       `json:"name"`
+	Model       string       `json:"model"`
+	RemoteModel string       `json:"remote_model,omitempty"`
+	RemoteHost  string       `json:"remote_host,omitempty"`
+	ModifiedAt  time.Time    `json:"modified_at"`
+	Size        int64        `json:"size"`
+	Digest      string       `json:"digest"`
+	Details     ModelDetails `json:"details,omitempty"`
 }
 
 // ProcessModelResponse is a single model description in [ProcessResponse].
@@ -537,6 +644,12 @@ type GenerateResponse struct {
 	// Model is the model name that generated the response.
 	Model string `json:"model"`
 
+	// RemoteModel is the name of the upstream model that generated the response.
+	RemoteModel string `json:"remote_model,omitempty"`
+
+	// RemoteHost is the URL of the upstream Ollama host that generated the response.
+	RemoteHost string `json:"remote_host,omitempty"`
+
 	// CreatedAt is the timestamp of the response.
 	CreatedAt time.Time `json:"created_at"`
 
@@ -560,6 +673,8 @@ type GenerateResponse struct {
 	Metrics
 
 	ToolCalls []ToolCall `json:"tool_calls,omitempty"`
+
+	DebugInfo *DebugInfo `json:"_debug_info,omitempty"`
 }
 
 // ModelDetails provides details about a model.
@@ -572,6 +687,18 @@ type ModelDetails struct {
 	QuantizationLevel string   `json:"quantization_level"`
 }
 
+// UserResponse provides information about a user.
+type UserResponse struct {
+	ID        uuid.UUID `json:"id"`
+	Email     string    `json:"email"`
+	Name      string    `json:"name"`
+	Bio       string    `json:"bio,omitempty"`
+	AvatarURL string    `json:"avatarurl,omitempty"`
+	FirstName string    `json:"firstname,omitempty"`
+	LastName  string    `json:"lastname,omitempty"`
+	Plan      string    `json:"plan,omitempty"`
+}
+
 // Tensor describes the metadata for a given tensor.
 type Tensor struct {
 	Name  string   `json:"name"`
@@ -769,8 +896,8 @@ func (t *ThinkValue) IsString() bool {
 	return ok
 }
 
-// AsBool returns the value as a bool (true if enabled in any way)
-func (t *ThinkValue) AsBool() bool {
+// Bool returns the value as a bool (true if enabled in any way)
+func (t *ThinkValue) Bool() bool {
 	if t == nil || t.Value == nil {
 		return false
 	}
@@ -786,8 +913,8 @@ func (t *ThinkValue) AsBool() bool {
 	}
 }
 
-// AsString returns the value as a string
-func (t *ThinkValue) AsString() string {
+// String returns the value as a string
+func (t *ThinkValue) String() string {
 	if t == nil || t.Value == nil {
 		return ""
 	}
@@ -825,7 +952,7 @@ func (t *ThinkValue) UnmarshalJSON(data []byte) error {
 		return nil
 	}
 
-	return fmt.Errorf("think must be a boolean or string (\"high\", \"medium\", \"low\")")
+	return fmt.Errorf("think must be a boolean or string (\"high\", \"medium\", \"low\", true, or false)")
 }
 
 // MarshalJSON implements json.Marshaler
@@ -860,7 +987,7 @@ func (d *Duration) UnmarshalJSON(b []byte) (err error) {
 		if t < 0 {
 			d.Duration = time.Duration(math.MaxInt64)
 		} else {
-			d.Duration = time.Duration(int(t) * int(time.Second))
+			d.Duration = time.Duration(t * float64(time.Second))
 		}
 	case string:
 		d.Duration, err = time.ParseDuration(t)