feat(config): per-provider temperature defaults and docs\n\n- Add , , to config with coding-friendly default 0.2.\n- Wire defaults through providers (OpenAI, Copilot, Ollama).\n- Update CLI and LSP runners to pass configured temperatures.\n- Document temperature behavior and examples in README.\n- Update config.json.example to show new keys.

11 files changed, 200 insertions, 90 deletions

diff --git a/IDEAS.md b/IDEAS.md
index 11cae34..a455e7e 100644
--- a/IDEAS.md
+++ b/IDEAS.md
@@ -15,7 +15,8 @@
 
 ### New features
 
-* [ ] Use hexai as a gh copilot... CLI replacemant for command line questions
+* [ ] implement a code action for selected code block the way via a unix pipe as faster access in helix
+* [x] Use hexai as a gh copilot... CLI replacemant for command line questions
 * [ ] Resolve diagnostics code action feature
 * [X] LSP server to be used with the Helix text editor
 * [X] Code completion using LLMs
diff --git a/README.md b/README.md
index bfa5cb9..cb74f95 100644
--- a/README.md
+++ b/README.md
@@ -4,6 +4,8 @@
 
 Hexai, the AI LSP for the Helix editor and also a simple command line tool to interact with LLMs in general.
 
+It has been coded with AI and human review.
+
 Hexai exposes a simple LLM provider interface. It supports OpenAI, GitHub Copilot, and a local Ollama server. Provider selection and models are configured via a JSON configuration file.
 
 ## Configuration
@@ -25,18 +27,21 @@ Hexai exposes a simple LLM provider interface. It supports OpenAI, GitHub Copilo
   "provider": "ollama",
   "copilot_model": "gpt-4.1",
   "copilot_base_url": "https://api.githubcopilot.com",
+  "copilot_temperature": 0.2,
   "openai_model": "gpt-4.1",
   "openai_base_url": "https://api.openai.com/v1",
-  "ollama_model": "qwen2.5-coder:latest",
-  "ollama_base_url": "http://localhost:11434"
+  "openai_temperature": 0.2,
+  "ollama_model": "qwen3-coder:30b-a3b-q4_K_M",
+  "ollama_base_url": "http://localhost:11434",
+  "ollama_temperature": 0.2
 }
 ```
 
 * context_mode: minimal | window | file-on-new-func | always-full
 * provider: openai | copilot | ollama
-* openai_model, openai_base_url: OpenAI-only options
-* copilot_model, copilot_base_url: Copilot-only options
-* ollama_model, ollama_base_url: Ollama-only options
+* openai_model, openai_base_url, openai_temperature: OpenAI-only options
+* copilot_model, copilot_base_url, copilot_temperature: Copilot-only options
+* ollama_model, ollama_base_url, ollama_temperature: Ollama-only options
 
 Ensure `OPENAI_API_KEY` or `COPILOT_API_KEY` is set in your environment according to your chosen provider.
 
@@ -51,6 +56,7 @@ Ensure `OPENAI_API_KEY` or `COPILOT_API_KEY` is set in your environment accordin
 - In config file:
   - `openai_model` — model name (default: `gpt-4.1`).
   - `openai_base_url` — API base (default: `https://api.openai.com/v1`).
+  - `openai_temperature` — default temperature (coding-friendly default `0.2`).
 
 ### GitHub Copilot configuration
 
@@ -58,15 +64,46 @@ Ensure `OPENAI_API_KEY` or `COPILOT_API_KEY` is set in your environment accordin
 - In config file:
   - `copilot_model` — model name (default: `gpt-4.1`).
   - `copilot_base_url` — API base (default: `https://api.githubcopilot.com`).
+  - `copilot_temperature` — default temperature (coding-friendly default `0.2`).
 
 ### Ollama configuration (local)
 
 - In config file:
-  - `ollama_model` — model name/tag (default: `qwen2.5-coder:latest`).
+  - `ollama_model` — model name/tag (default: `qwen3-coder:30b-a3b-q4_K_M`).
   - `ollama_base_url` — base URL to Ollama (default: `http://localhost:11434`).
+  - `ollama_temperature` — default temperature (coding-friendly default `0.2`).
+
+### Temperature behavior
+
+* What it is: Temperature controls how random/creative the model's word choices are.
+  Lower values (≈0–0.3) are more deterministic and precise; higher values (≈0.7+)
+  produce more diverse, creative outputs.
+* Default for coding: When not specified in the config, Hexai uses a
+  coding-friendly default temperature of `0.2` for all providers.
+* Per-provider override: Set `openai_temperature`, `copilot_temperature`, or
+  `ollama_temperature` to override. Valid ranges depend on the provider, but
+  typically `0.0`–`2.0`.
+* LSP vs CLI: The LSP sometimes overrides temperature for specific actions
+  (e.g., `0.1`–`0.2` for completions). The CLI uses the configured provider
+  default unless you change it.
+
+Recommended ranges and use cases:
+
+- 0.0–0.3: Deterministic, precise, minimal tangents. Best for code
+  refactoring, bug fixes, tests, and data extraction.
+- 0.4–0.7: Balanced creativity and coherence. General Q&A and most writing.
+- 0.8–1.2+: Highly creative/varied. Brainstorming, fiction, or ad copy; may
+  increase risk of off-target or verbose outputs.
+
+Guidance:
+
+- Lower temperature increases consistency and predictability, but can repeat
+  or be terse.
+- Higher temperature increases diversity of phrasing and ideas, but can wander
+  or introduce mistakes.
 
 Notes:
-- For Ollama, ensure the model is available locally (e.g., `ollama pull qwen2.5-coder:latest`).
+- For Ollama, ensure the model is available locally (e.g., `ollama pull qwen3-coder:30b-a3b-q4_K_M`).
 - If you run Ollama in OpenAI‑compatible mode, you may alternatively use the
   OpenAI provider with `openai_base_url` in the config pointing to your local endpoint.
 
@@ -104,11 +141,11 @@ Note, that we have also configured other LSPs here (for Go, `gopls` and `golangc
 Hexai LSP supports inline trigger tags you can type in your code to request an
 action from the LLM and then clean up the tag automatically.
 
-- `;some prompt here;`: Do what is written in `some prompt text here`, then remove just the prompt.
+- ``: Do what is written in `some prompt text here`, then remove just the prompt.
   - Strict form: no space after the first ``.
   - An optional single space immediately after the closing `;` is also removed.
 - Spaced variants such as `; text ; spaced ;` are ignored.
-- `some text here ;;some prompt;`
+
 
 ## Code actions
 
@@ -156,4 +193,3 @@ hexai 'install ripgrep on macOS'
 # Verbose explanation
 hexai 'install ripgrep on macOS and explain'
 ```
-
diff --git a/config.json.example b/config.json.example
index 359e862..04ecfb7 100644
--- a/config.json.example
+++ b/config.json.example
@@ -11,10 +11,13 @@
 
   "openai_model": "gpt-4.1",
   "openai_base_url": "https://api.openai.com/v1",
+  "openai_temperature": 0.2,
 
-  "ollama_model": "qwen2.5-coder:latest",
+  "ollama_model": "qwen3-coder:30b-a3b-q4_K_M",
   "ollama_base_url": "http://localhost:11434",
+  "ollama_temperature": 0.2,
 
   "copilot_model": "gpt-4.1",
-  "copilot_base_url": "https://api.githubcopilot.com"
+  "copilot_base_url": "https://api.githubcopilot.com",
+  "copilot_temperature": 0.2
 }
diff --git a/internal/appconfig/config.go b/internal/appconfig/config.go
index 7027547..c377467 100644
--- a/internal/appconfig/config.go
+++ b/internal/appconfig/config.go
@@ -13,7 +13,7 @@ import (
 
 // App holds user-configurable settings read from ~/.config/hexai/config.json.
 type App struct {
-	MaxTokens          int    `json:"max_tokens"`
+    MaxTokens          int    `json:"max_tokens"`
 	ContextMode        string `json:"context_mode"`
 	ContextWindowLines int    `json:"context_window_lines"`
 	MaxContextTokens   int    `json:"max_context_tokens"`
@@ -22,23 +22,35 @@ type App struct {
 	TriggerCharacters []string `json:"trigger_characters"`
 	Provider          string   `json:"provider"`
 
-	// Provider-specific options
-	OpenAIBaseURL  string `json:"openai_base_url"`
-	OpenAIModel    string `json:"openai_model"`
-	OllamaBaseURL  string `json:"ollama_base_url"`
-	OllamaModel    string `json:"ollama_model"`
-	CopilotBaseURL string `json:"copilot_base_url"`
-	CopilotModel   string `json:"copilot_model"`
+    // Provider-specific options
+    OpenAIBaseURL  string `json:"openai_base_url"`
+    OpenAIModel    string `json:"openai_model"`
+    // Default temperature for OpenAI requests (nil means use provider default)
+    OpenAITemperature *float64 `json:"openai_temperature"`
+    OllamaBaseURL  string `json:"ollama_base_url"`
+    OllamaModel    string `json:"ollama_model"`
+    // Default temperature for Ollama requests (nil means use provider default)
+    OllamaTemperature *float64 `json:"ollama_temperature"`
+    CopilotBaseURL string `json:"copilot_base_url"`
+    CopilotModel   string `json:"copilot_model"`
+    // Default temperature for Copilot requests (nil means use provider default)
+    CopilotTemperature *float64 `json:"copilot_temperature"`
 }
 
 // Constructor: defaults for App (kept first among functions)
 func newDefaultConfig() App {
+    // Coding-friendly default temperature across providers
+    // Users can override per provider in config.json (including 0.0).
+    t := 0.2
     return App{
         MaxTokens:          4000,
         ContextMode:        "always-full",
         ContextWindowLines: 120,
         MaxContextTokens:   4000,
         LogPreviewLimit:    100,
+        OpenAITemperature:  &t,
+        OllamaTemperature:  &t,
+        CopilotTemperature: &t,
     }
 }
 
@@ -115,18 +127,27 @@ func (a *App) mergeWith(other *App) {
     if strings.TrimSpace(other.OpenAIModel) != "" {
         a.OpenAIModel = other.OpenAIModel
     }
+    if other.OpenAITemperature != nil { // allow explicit 0.0
+        a.OpenAITemperature = other.OpenAITemperature
+    }
     if strings.TrimSpace(other.OllamaBaseURL) != "" {
         a.OllamaBaseURL = other.OllamaBaseURL
     }
     if strings.TrimSpace(other.OllamaModel) != "" {
         a.OllamaModel = other.OllamaModel
     }
+    if other.OllamaTemperature != nil { // allow explicit 0.0
+        a.OllamaTemperature = other.OllamaTemperature
+    }
     if strings.TrimSpace(other.CopilotBaseURL) != "" {
         a.CopilotBaseURL = other.CopilotBaseURL
     }
     if strings.TrimSpace(other.CopilotModel) != "" {
         a.CopilotModel = other.CopilotModel
     }
+    if other.CopilotTemperature != nil { // allow explicit 0.0
+        a.CopilotTemperature = other.CopilotTemperature
+    }
 }
 
 func getConfigPath() (string, error) {
diff --git a/internal/hexaicli/run.go b/internal/hexaicli/run.go
index 11e64b3..839daef 100644
--- a/internal/hexaicli/run.go
+++ b/internal/hexaicli/run.go
@@ -68,15 +68,18 @@ func readInput(stdin io.Reader, args []string) (string, error) {
 
 // newClientFromConfig builds an LLM client from the app config and env keys.
 func newClientFromConfig(cfg appconfig.App) (llm.Client, error) {
-	llmCfg := llm.Config{
-		Provider:       cfg.Provider,
-		OpenAIBaseURL:  cfg.OpenAIBaseURL,
-		OpenAIModel:    cfg.OpenAIModel,
-		OllamaBaseURL:  cfg.OllamaBaseURL,
-		OllamaModel:    cfg.OllamaModel,
-		CopilotBaseURL: cfg.CopilotBaseURL,
-		CopilotModel:   cfg.CopilotModel,
-	}
+    llmCfg := llm.Config{
+        Provider:       cfg.Provider,
+        OpenAIBaseURL:  cfg.OpenAIBaseURL,
+        OpenAIModel:    cfg.OpenAIModel,
+        OpenAITemperature: cfg.OpenAITemperature,
+        OllamaBaseURL:  cfg.OllamaBaseURL,
+        OllamaModel:    cfg.OllamaModel,
+        OllamaTemperature: cfg.OllamaTemperature,
+        CopilotBaseURL: cfg.CopilotBaseURL,
+        CopilotModel:   cfg.CopilotModel,
+        CopilotTemperature: cfg.CopilotTemperature,
+    }
 	oaKey := os.Getenv("OPENAI_API_KEY")
 	cpKey := os.Getenv("COPILOT_API_KEY")
 	return llm.NewFromConfig(llmCfg, oaKey, cpKey)
diff --git a/internal/hexailsp/run.go b/internal/hexailsp/run.go
index 8a79b12..2eee3aa 100644
--- a/internal/hexailsp/run.go
+++ b/internal/hexailsp/run.go
@@ -52,10 +52,13 @@ func RunWithFactory(logPath string, stdin io.Reader, stdout io.Writer, logger *l
 			Provider:       cfg.Provider,
 			OpenAIBaseURL:  cfg.OpenAIBaseURL,
 			OpenAIModel:    cfg.OpenAIModel,
+			OpenAITemperature: cfg.OpenAITemperature,
 			OllamaBaseURL:  cfg.OllamaBaseURL,
 			OllamaModel:    cfg.OllamaModel,
+			OllamaTemperature: cfg.OllamaTemperature,
 			CopilotBaseURL: cfg.CopilotBaseURL,
 			CopilotModel:   cfg.CopilotModel,
+			CopilotTemperature: cfg.CopilotTemperature,
 		}
 		oaKey := os.Getenv("OPENAI_API_KEY")
 		cpKey := os.Getenv("COPILOT_API_KEY")
diff --git a/internal/llm/copilot.go b/internal/llm/copilot.go
index 680e7ec..47ce11e 100644
--- a/internal/llm/copilot.go
+++ b/internal/llm/copilot.go
@@ -22,6 +22,7 @@ type copilotClient struct {
     baseURL      string
     defaultModel string
     chatLogger   logging.ChatLogger
+    defaultTemperature *float64
 }
 
 type copilotChatRequest struct {
@@ -55,7 +56,7 @@ type copilotChatResponse struct {
 }
 
 // Constructor (kept among the first functions by convention)
-func newCopilot(baseURL, model, apiKey string) Client {
+func newCopilot(baseURL, model, apiKey string, defaultTemp *float64) Client {
     if strings.TrimSpace(baseURL) == "" {
         baseURL = "https://api.githubcopilot.com"
     }
@@ -68,6 +69,7 @@ func newCopilot(baseURL, model, apiKey string) Client {
         baseURL:      strings.TrimRight(baseURL, "/"),
         defaultModel: model,
         chatLogger:   logging.NewChatLogger("copilot"),
+        defaultTemperature: defaultTemp,
     }
 }
 
@@ -101,9 +103,12 @@ func (c copilotClient) Chat(ctx context.Context, messages []Message, opts ...Req
 	for i, m := range messages {
 		req.Messages[i] = copilotMessage{Role: m.Role, Content: m.Content}
 	}
-	if o.Temperature != 0 {
-		req.Temperature = &o.Temperature
-	}
+    if o.Temperature != 0 {
+        req.Temperature = &o.Temperature
+    } else if c.defaultTemperature != nil {
+        t := *c.defaultTemperature
+        req.Temperature = &t
+    }
 	if o.MaxTokens > 0 {
 		req.MaxTokens = &o.MaxTokens
 	}
diff --git a/internal/llm/ollama.go b/internal/llm/ollama.go
index 14aa558..20dfe2a 100644
--- a/internal/llm/ollama.go
+++ b/internal/llm/ollama.go
@@ -22,13 +22,14 @@ type ollamaClient struct {
     baseURL      string
     defaultModel string
     chatLogger   logging.ChatLogger
+    defaultTemperature *float64
 }
 
 type ollamaChatRequest struct {
-    Model    string      `json:"model"`
-    Messages []oaMessage `json:"messages"`
-    Stream   bool        `json:"stream"`
-    Options  any         `json:"options,omitempty"`
+	Model    string      `json:"model"`
+	Messages []oaMessage `json:"messages"`
+	Stream   bool        `json:"stream"`
+	Options  any         `json:"options,omitempty"`
 }
 
 type ollamaChatResponse struct {
@@ -41,21 +42,23 @@ type ollamaChatResponse struct {
 }
 
 // Constructor (kept among the first functions by convention)
-func newOllama(baseURL, model string) Client {
-    if strings.TrimSpace(baseURL) == "" {
-        baseURL = "http://localhost:11434"
-    }
-    if strings.TrimSpace(model) == "" {
-        model = "qwen2.5-coder:latest"
-    }
+func newOllama(baseURL, model string, defaultTemp *float64) Client {
+	if strings.TrimSpace(baseURL) == "" {
+		baseURL = "http://localhost:11434"
+	}
+	if strings.TrimSpace(model) == "" {
+		model = "qwen3-coder:30b-a3b-q4_K_M`"
+	}
     return ollamaClient{
         httpClient:   &http.Client{Timeout: 30 * time.Second},
         baseURL:      strings.TrimRight(baseURL, "/"),
         defaultModel: model,
         chatLogger:   logging.NewChatLogger("ollama"),
+        defaultTemperature: defaultTemp,
     }
 }
 
+// TODO: This function is too long and should be refactored for readability and maintainability.
 func (c ollamaClient) Chat(ctx context.Context, messages []Message, opts ...RequestOption) (string, error) {
 	o := Options{Model: c.defaultModel}
 	for _, opt := range opts {
@@ -86,9 +89,11 @@ func (c ollamaClient) Chat(ctx context.Context, messages []Message, opts ...Requ
 
 	// Build options map only if any option is set
 	optsMap := map[string]any{}
-	if o.Temperature != 0 {
-		optsMap["temperature"] = o.Temperature
-	}
+    if o.Temperature != 0 {
+        optsMap["temperature"] = o.Temperature
+    } else if c.defaultTemperature != nil {
+        optsMap["temperature"] = *c.defaultTemperature
+    }
 	if o.MaxTokens > 0 {
 		optsMap["num_predict"] = o.MaxTokens
 	}
@@ -177,9 +182,11 @@ func (c ollamaClient) ChatStream(ctx context.Context, messages []Message, onDelt
 	}
 	// Build options map
 	optsMap := map[string]any{}
-	if o.Temperature != 0 {
-		optsMap["temperature"] = o.Temperature
-	}
+    if o.Temperature != 0 {
+        optsMap["temperature"] = o.Temperature
+    } else if c.defaultTemperature != nil {
+        optsMap["temperature"] = *c.defaultTemperature
+    }
 	if o.MaxTokens > 0 {
 		optsMap["num_predict"] = o.MaxTokens
 	}
@@ -241,6 +248,6 @@ func (c ollamaClient) ChatStream(ctx context.Context, messages []Message, onDelt
 			break
 		}
 	}
-    logging.Logf("llm/ollama ", "stream end duration=%s", time.Since(start))
-    return nil
+	logging.Logf("llm/ollama ", "stream end duration=%s", time.Since(start))
+	return nil
 }
diff --git a/internal/llm/openai.go b/internal/llm/openai.go
index 8dc2907..5348def 100644
--- a/internal/llm/openai.go
+++ b/internal/llm/openai.go
@@ -23,6 +23,7 @@ type openAIClient struct {
     baseURL      string
     defaultModel string
     chatLogger   logging.ChatLogger
+    defaultTemperature *float64
 }
 
 type oaChatRequest struct {
@@ -75,7 +76,7 @@ type oaStreamChunk struct {
 // Constructor (kept among the first functions by convention)
 // newOpenAI constructs an OpenAI client using explicit configuration values.
 // The apiKey may be empty; calls will fail until a valid key is supplied.
-func newOpenAI(baseURL, model, apiKey string) Client {
+func newOpenAI(baseURL, model, apiKey string, defaultTemp *float64) Client {
     if strings.TrimSpace(baseURL) == "" {
         baseURL = "https://api.openai.com/v1"
     }
@@ -88,6 +89,7 @@ func newOpenAI(baseURL, model, apiKey string) Client {
         baseURL:      baseURL,
         defaultModel: model,
         chatLogger:   logging.NewChatLogger("openai"),
+        defaultTemperature: defaultTemp,
     }
 }
 
@@ -120,9 +122,13 @@ func (c openAIClient) Chat(ctx context.Context, messages []Message, opts ...Requ
 	for i, m := range messages {
 		req.Messages[i] = oaMessage{Role: m.Role, Content: m.Content}
 	}
-	if o.Temperature != 0 {
-		req.Temperature = &o.Temperature
-	}
+    // Decide temperature: request option overrides config default.
+    if o.Temperature != 0 {
+        req.Temperature = &o.Temperature
+    } else if c.defaultTemperature != nil {
+        t := *c.defaultTemperature
+        req.Temperature = &t
+    }
 	if o.MaxTokens > 0 {
 		req.MaxTokens = &o.MaxTokens
 	}
@@ -212,9 +218,12 @@ func (c openAIClient) ChatStream(ctx context.Context, messages []Message, onDelt
 	for i, m := range messages {
 		req.Messages[i] = oaMessage{Role: m.Role, Content: m.Content}
 	}
-	if o.Temperature != 0 {
-		req.Temperature = &o.Temperature
-	}
+    if o.Temperature != 0 {
+        req.Temperature = &o.Temperature
+    } else if c.defaultTemperature != nil {
+        t := *c.defaultTemperature
+        req.Temperature = &t
+    }
 	if o.MaxTokens > 0 {
 		req.MaxTokens = &o.MaxTokens
 	}
diff --git a/internal/llm/provider.go b/internal/llm/provider.go
index c605081..ed9ca59 100644
--- a/internal/llm/provider.go
+++ b/internal/llm/provider.go
@@ -54,40 +54,56 @@ func WithStop(stop ...string) RequestOption {
 
 // Config defines provider configuration read from the Hexai config file.
 type Config struct {
-	Provider string
-	// OpenAI options
-	OpenAIBaseURL string
-	OpenAIModel   string
-	// Ollama options
-	OllamaBaseURL string
-	OllamaModel   string
-	// Copilot options
-	CopilotBaseURL string
-	CopilotModel   string
+    Provider string
+    // OpenAI options
+    OpenAIBaseURL string
+    OpenAIModel   string
+    OpenAITemperature *float64
+    // Ollama options
+    OllamaBaseURL string
+    OllamaModel   string
+    OllamaTemperature *float64
+    // Copilot options
+    CopilotBaseURL string
+    CopilotModel   string
+    CopilotTemperature *float64
 }
 
 // NewFromConfig creates an LLM client using only the supplied configuration.
 // The OpenAI API key is supplied separately and may be read from the environment
 // by the caller; other environment-based configuration is not used.
 func NewFromConfig(cfg Config, openAIAPIKey, copilotAPIKey string) (Client, error) {
-	p := strings.ToLower(strings.TrimSpace(cfg.Provider))
-	if p == "" {
-		p = "openai"
-	}
-	switch p {
-	case "openai":
-		if strings.TrimSpace(openAIAPIKey) == "" {
-			return nil, errors.New("missing OPENAI_API_KEY for provider openai")
-		}
-		return newOpenAI(cfg.OpenAIBaseURL, cfg.OpenAIModel, openAIAPIKey), nil
-	case "ollama":
-		return newOllama(cfg.OllamaBaseURL, cfg.OllamaModel), nil
-	case "copilot":
-		if strings.TrimSpace(copilotAPIKey) == "" {
-			return nil, errors.New("missing COPILOT_API_KEY for provider copilot")
-		}
-		return newCopilot(cfg.CopilotBaseURL, cfg.CopilotModel, copilotAPIKey), nil
-	default:
-		return nil, errors.New("unknown LLM provider: " + p)
-	}
+    p := strings.ToLower(strings.TrimSpace(cfg.Provider))
+    if p == "" {
+        p = "openai"
+    }
+    switch p {
+    case "openai":
+        if strings.TrimSpace(openAIAPIKey) == "" {
+            return nil, errors.New("missing OPENAI_API_KEY for provider openai")
+        }
+        // Set coding-friendly default temperature if none provided
+        if cfg.OpenAITemperature == nil {
+            t := 0.2
+            cfg.OpenAITemperature = &t
+        }
+        return newOpenAI(cfg.OpenAIBaseURL, cfg.OpenAIModel, openAIAPIKey, cfg.OpenAITemperature), nil
+    case "ollama":
+        if cfg.OllamaTemperature == nil {
+            t := 0.2
+            cfg.OllamaTemperature = &t
+        }
+        return newOllama(cfg.OllamaBaseURL, cfg.OllamaModel, cfg.OllamaTemperature), nil
+    case "copilot":
+        if strings.TrimSpace(copilotAPIKey) == "" {
+            return nil, errors.New("missing COPILOT_API_KEY for provider copilot")
+        }
+        if cfg.CopilotTemperature == nil {
+            t := 0.2
+            cfg.CopilotTemperature = &t
+        }
+        return newCopilot(cfg.CopilotBaseURL, cfg.CopilotModel, copilotAPIKey, cfg.CopilotTemperature), nil
+    default:
+        return nil, errors.New("unknown LLM provider: " + p)
+    }
 }
diff --git a/internal/llm/util.go b/internal/llm/util.go
new file mode 100644
index 0000000..b99d7c8
--- /dev/null
+++ b/internal/llm/util.go
@@ -0,0 +1,6 @@
+package llm
+
+import "errors"
+
+// small helper to keep return type consistent
+func nilStringErr(msg string) (string, error) { return "", errors.New(msg) }