feat: enable per-provider SSE streaming via model_list stream field

Wire ModelConfig.Stream (*bool) through to openai_compat.WithStream
so each model_list entry can opt in to SSE streaming independently.

- Add NewHTTPProviderFromConfig that reads all ModelConfig fields
  including stream
- Implement StreamingProvider interface (CanStream/ChatStream) on
  HTTPProvider by delegating to openai_compat.Provider
- Switch factory_provider.go (openai, openai-compat, anthropic
  API-key paths) to use NewHTTPProviderFromConfig

Usage: set "stream": true in a model_list entry to enable streaming.
Default behavior (no stream field) is unchanged.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

pkg/providers/factory_provider.go

@@ -87,13 +87,7 @@ func CreateProviderFromConfig(cfg *config.ModelConfig) (LLMProvider, string, err
 		if apiBase == "" {
 			apiBase = getDefaultAPIBase(protocol)
 		}
-		return NewHTTPProviderWithMaxTokensFieldAndRequestTimeout(
-			cfg.APIKey,
-			apiBase,
-			cfg.Proxy,
-			cfg.MaxTokensField,
-			cfg.RequestTimeout,
-		), modelID, nil
+		return NewHTTPProviderFromConfig(cfg, apiBase), modelID, nil
 
 	case "azure", "azure-openai":
 		// Azure OpenAI uses deployment-based URLs, api-key header auth,
@@ -126,13 +120,7 @@ func CreateProviderFromConfig(cfg *config.ModelConfig) (LLMProvider, string, err
 		if apiBase == "" {
 			apiBase = getDefaultAPIBase(protocol)
 		}
-		return NewHTTPProviderWithMaxTokensFieldAndRequestTimeout(
-			cfg.APIKey,
-			apiBase,
-			cfg.Proxy,
-			cfg.MaxTokensField,
-			cfg.RequestTimeout,
-		), modelID, nil
+		return NewHTTPProviderFromConfig(cfg, apiBase), modelID, nil
 
 	case "anthropic":
 		if cfg.AuthMethod == "oauth" || cfg.AuthMethod == "token" {
@@ -151,13 +139,7 @@ func CreateProviderFromConfig(cfg *config.ModelConfig) (LLMProvider, string, err
 		if cfg.APIKey == "" {
 			return nil, "", fmt.Errorf("api_key is required for anthropic protocol (model: %s)", cfg.Model)
 		}
-		return NewHTTPProviderWithMaxTokensFieldAndRequestTimeout(
-			cfg.APIKey,
-			apiBase,
-			cfg.Proxy,
-			cfg.MaxTokensField,
-			cfg.RequestTimeout,
-		), modelID, nil
+		return NewHTTPProviderFromConfig(cfg, apiBase), modelID, nil
 
 	case "anthropic-messages":
 		// Anthropic Messages API with native format (HTTP-based, no SDK)

pkg/providers/http_provider.go

@@ -10,7 +10,9 @@ import (
 	"context"
 	"time"
 
+	"github.com/sipeed/picoclaw/pkg/config"
 	"github.com/sipeed/picoclaw/pkg/providers/openai_compat"
+	"github.com/sipeed/picoclaw/pkg/providers/protocoltypes"
 )
 
 type HTTPProvider struct {
@@ -42,6 +44,21 @@ func NewHTTPProviderWithMaxTokensFieldAndRequestTimeout(
 	}
 }
 
+// NewHTTPProviderFromConfig creates an HTTPProvider from a ModelConfig,
+// honoring all optional fields including stream.
+func NewHTTPProviderFromConfig(cfg *config.ModelConfig, apiBase string) *HTTPProvider {
+	opts := []openai_compat.Option{
+		openai_compat.WithMaxTokensField(cfg.MaxTokensField),
+		openai_compat.WithRequestTimeout(time.Duration(cfg.RequestTimeout) * time.Second),
+	}
+	if cfg.Stream != nil && *cfg.Stream {
+		opts = append(opts, openai_compat.WithStream(true))
+	}
+	return &HTTPProvider{
+		delegate: openai_compat.NewProvider(cfg.APIKey, apiBase, cfg.Proxy, opts...),
+	}
+}
+
 func (p *HTTPProvider) Chat(
 	ctx context.Context,
 	messages []Message,
@@ -52,6 +69,22 @@ func (p *HTTPProvider) Chat(
 	return p.delegate.Chat(ctx, messages, tools, model, options)
 }
 
+// CanStream returns true when SSE streaming is enabled.
+func (p *HTTPProvider) CanStream() bool {
+	return p.delegate.CanStream()
+}
+
+// ChatStream opens an SSE connection and returns a channel of StreamEvent.
+func (p *HTTPProvider) ChatStream(
+	ctx context.Context,
+	messages []Message,
+	tools []ToolDefinition,
+	model string,
+	options map[string]any,
+) (<-chan protocoltypes.StreamEvent, error) {
+	return p.delegate.ChatStream(ctx, messages, tools, model, options)
+}
+
 func (p *HTTPProvider) GetDefaultModel() string {
 	return ""
 }