logging: migrate LSP logs to global singleton (internal/logging); use consistent colors/prefix; refactor LLM provider to use global logger and remove per-client loggers

8 files changed, 100 insertions, 92 deletions

diff --git a/README.md b/README.md
index 22a7821..c3c77c2 100644
--- a/README.md
+++ b/README.md
@@ -12,7 +12,7 @@ Hexai exposes a simple LLM provider interface and uses OpenAI by default for
 code completion when `OPENAI_API_KEY` is present in the environment.
 
 - Required: set `OPENAI_API_KEY` to your OpenAI API key.
-- Optional: set `OPENAI_MODEL` (default: `gpt-4o-mini`).
+- Optional: set `OPENAI_MODEL`.
 - Optional: set `OPENAI_BASE_URL` to point at a compatible endpoint.
 
 If no key is configured, Hexai will fall back to a basic, local completion.
@@ -20,17 +20,17 @@ If no key is configured, Hexai will fall back to a basic, local completion.
 ## CLI usage and configuration
 
 - Run LSP server over stdio:
-  - `hexai -stdio`
+  - `hexai`
 
 - Completion settings:
-  - `-max-tokens`: maximum tokens for LLM completions (default `500`). If the flag isn’t provided, `HEXAI_MAX_TOKENS` is used when set.
+  - `-max-tokens`: maximum tokens for LLM completions. If the flag isn’t provided, `HEXAI_MAX_TOKENS` is used when set.
   - `-context-mode`: how much additional context to include with completion prompts (If the flag isn’t provided, `HEXAI_CONTEXT_MODE` is used when set). One of:
     - `minimal`: no extra context
     - `window`: include a sliding window around the cursor
-    - `file-on-new-func` (default): include the full file only when defining a new function (cursor before the opening `{`)
+    - `file-on-new-func`ude the full file only when defining a new function (cursor before the opening `{`)
     - `always-full`: always include the full file (may be slower/costly)
-  - `-context-window-lines`: line count for the sliding window when `context-mode=window` (default `120`).
-  - `-max-context-tokens`: budget for additional context tokens (default `2000`). If the flag isn’t provided, `HEXAI_MAX_CONTEXT_TOKENS` is used when set.
+  - `-context-window-lines`: line count for the sliding window when `context-mode=window`.
+  - `-max-context-tokens`: budget for additional context tokens. If the flag isn’t provided, `HEXAI_MAX_CONTEXT_TOKENS` is used when set.
 
 Notes:
 - Token estimation for truncation uses a simple 4 chars/token heuristic.
@@ -38,13 +38,13 @@ Notes:
 
 ### Flags quick reference
 
-| Flag                    | Default            | Env override               | Description                                        |
-|-------------------------|--------------------|----------------------------|----------------------------------------------------|
-| `-stdio`                | `true`             | —                          | Run as LSP over stdio (only supported mode).       |
-| `-log`                  | `/tmp/hexai.log`   | —                          | Path to log file (optional).                       |
-| `-max-tokens`           | `500`              | `HEXAI_MAX_TOKENS`         | Max tokens for LLM completions.                    |
-| `-context-mode`         | `file-on-new-func` | `HEXAI_CONTEXT_MODE`       | `minimal` `window` `file-on-new-func` `always-full` |
-| `-context-window-lines` | `120`              | `HEXAI_CONTEXT_WINDOW_LINES` | Lines around cursor when using `window` mode.      |
-| `-max-context-tokens`   | `2000`             | `HEXAI_MAX_CONTEXT_TOKENS` | Token budget for additional context.               |
-| `-log-preview-limit`    | `0` (unlimited)    | `HEXAI_LOG_PREVIEW_LIMIT`  | Limit characters shown in LLM preview logs.        |
-| `-no-disk-io`           | `true`             | `HEXAI_NO_DISK_IO`         | Disallow any disk reads for context.               |
+| Flag                    | Env override               | Description                                        |
+|-------------------------|----------------------------|----------------------------------------------------|
+| `-stdio`                | —                          | Run as LSP over stdio (only supported mode).       |
+| `-log`                  | —                          | Path to log file (optional).                       |
+| `-max-tokens`           | `HEXAI_MAX_TOKENS`         | Max tokens for LLM completions.                    |
+| `-context-mode`         | `HEXAI_CONTEXT_MODE`       | `minimal` `window` `file-on-new-func` `always-full` |
+| `-context-window-lines` | `HEXAI_CONTEXT_WINDOW_LINES` | Lines around cursor when using `window` mode.      |
+| `-max-context-tokens`   | `HEXAI_MAX_CONTEXT_TOKENS` | Token budget for additional context.               |
+| `-log-preview-limit`    | `HEXAI_LOG_PREVIEW_LIMIT`  | Limit characters shown in LLM preview logs.        |
+| `-no-disk-io`           | `HEXAI_NO_DISK_IO`         | Disallow any disk reads for context.               |
diff --git a/internal/llm/openai.go b/internal/llm/openai.go
index 9b48782..279eca4 100644
--- a/internal/llm/openai.go
+++ b/internal/llm/openai.go
@@ -6,10 +6,11 @@ import (
 	"encoding/json"
 	"errors"
 	"fmt"
-	"log"
 	"net/http"
 	"os"
 	"time"
+
+	"hexai/internal/logging"
 )
 
 // openAIClient implements Client against OpenAI's Chat Completions API.
@@ -18,12 +19,11 @@ type openAIClient struct {
 	apiKey       string
 	baseURL      string
 	defaultModel string
-	logger       *log.Logger
 }
 
 // Colors and base styling are provided by logging.go
 
-func newOpenAIFromEnv(apiKey string, logger *log.Logger) Client {
+func newOpenAIFromEnv(apiKey string) Client {
 	base := os.Getenv("OPENAI_BASE_URL")
 	if base == "" {
 		base = "https://api.openai.com/v1"
@@ -37,7 +37,6 @@ func newOpenAIFromEnv(apiKey string, logger *log.Logger) Client {
 		apiKey:       apiKey,
 		baseURL:      base,
 		defaultModel: model,
-		logger:       logger,
 	}
 }
 
@@ -83,10 +82,10 @@ func (c *openAIClient) Chat(ctx context.Context, messages []Message, opts ...Req
 		o.Model = c.defaultModel
 	}
 	start := time.Now()
-    LogPrintf(c.logger, "llm/openai ", "chat start model=%s temp=%.2f max_tokens=%d stop=%d messages=%d", o.Model, o.Temperature, o.MaxTokens, len(o.Stop), len(messages))
+    logging.Logf("llm/openai ", "chat start model=%s temp=%.2f max_tokens=%d stop=%d messages=%d", o.Model, o.Temperature, o.MaxTokens, len(o.Stop), len(messages))
 	for i, m := range messages {
         // Sending context (cyan)
-        LogPrintf(c.logger, "llm/openai ", "msg[%d] role=%s size=%d preview=%s%s%s", i, m.Role, len(m.Content), AnsiCyan, previewForLog(m.Content), AnsiBase)
+        logging.Logf("llm/openai ", "msg[%d] role=%s size=%d preview=%s%s%s", i, m.Role, len(m.Content), logging.AnsiCyan, logging.PreviewForLog(m.Content), logging.AnsiBase)
 	}
 	req := oaChatRequest{Model: o.Model}
 	req.Messages = make([]oaMessage, len(messages))
@@ -109,7 +108,7 @@ func (c *openAIClient) Chat(ctx context.Context, messages []Message, opts ...Req
 		return "", err
 	}
 	endpoint := c.baseURL + "/chat/completions"
-    LogPrintf(c.logger, "llm/openai ", "POST %s", endpoint)
+    logging.Logf("llm/openai ", "POST %s", endpoint)
 	httpReq, err := http.NewRequestWithContext(ctx, http.MethodPost, endpoint, bytes.NewReader(body))
 	if err != nil {
 		c.logf("new request error: %v", err)
@@ -120,7 +119,7 @@ func (c *openAIClient) Chat(ctx context.Context, messages []Message, opts ...Req
 
 	resp, err := c.httpClient.Do(httpReq)
     if err != nil {
-        LogPrintf(c.logger, "llm/openai ", "%shttp error after %s: %v%s", AnsiRed, time.Since(start), err, AnsiBase)
+        logging.Logf("llm/openai ", "%shttp error after %s: %v%s", logging.AnsiRed, time.Since(start), err, logging.AnsiBase)
         return "", err
     }
 	defer resp.Body.Close()
@@ -128,31 +127,31 @@ func (c *openAIClient) Chat(ctx context.Context, messages []Message, opts ...Req
 		var apiErr oaChatResponse
 		_ = json.NewDecoder(resp.Body).Decode(&apiErr)
         if apiErr.Error != nil && apiErr.Error.Message != "" {
-            LogPrintf(c.logger, "llm/openai ", "%sapi error status=%d type=%s msg=%s duration=%s%s", AnsiRed, resp.StatusCode, apiErr.Error.Type, apiErr.Error.Message, time.Since(start), AnsiBase)
+            logging.Logf("llm/openai ", "%sapi error status=%d type=%s msg=%s duration=%s%s", logging.AnsiRed, resp.StatusCode, apiErr.Error.Type, apiErr.Error.Message, time.Since(start), logging.AnsiBase)
             return "", fmt.Errorf("openai error: %s (status %d)", apiErr.Error.Message, resp.StatusCode)
         }
-        LogPrintf(c.logger, "llm/openai ", "%shttp non-2xx status=%d duration=%s%s", AnsiRed, resp.StatusCode, time.Since(start), AnsiBase)
+        logging.Logf("llm/openai ", "%shttp non-2xx status=%d duration=%s%s", logging.AnsiRed, resp.StatusCode, time.Since(start), logging.AnsiBase)
         return "", fmt.Errorf("openai http error: status %d", resp.StatusCode)
     }
 	var out oaChatResponse
     if err := json.NewDecoder(resp.Body).Decode(&out); err != nil {
-        LogPrintf(c.logger, "llm/openai ", "%sdecode error after %s: %v%s", AnsiRed, time.Since(start), err, AnsiBase)
+        logging.Logf("llm/openai ", "%sdecode error after %s: %v%s", logging.AnsiRed, time.Since(start), err, logging.AnsiBase)
         return "", err
     }
     if len(out.Choices) == 0 {
-        LogPrintf(c.logger, "llm/openai ", "%sno choices returned duration=%s%s", AnsiRed, time.Since(start), AnsiBase)
+        logging.Logf("llm/openai ", "%sno choices returned duration=%s%s", logging.AnsiRed, time.Since(start), logging.AnsiBase)
         return "", errors.New("openai: no choices returned")
     }
     content := out.Choices[0].Message.Content
     // Received context (green)
-    LogPrintf(c.logger, "llm/openai ", "success choice=0 finish=%s size=%d preview=%s%s%s duration=%s", out.Choices[0].FinishReason, len(content), AnsiGreen, previewForLog(content), AnsiBase, time.Since(start))
+    logging.Logf("llm/openai ", "success choice=0 finish=%s size=%d preview=%s%s%s duration=%s", out.Choices[0].FinishReason, len(content), logging.AnsiGreen, logging.PreviewForLog(content), logging.AnsiBase, time.Since(start))
     return content, nil
 }
 
 // small helper to keep return type consistent
 func nilStringErr(msg string) (string, error) { return "", errors.New(msg) }
 
-func (c *openAIClient) logf(format string, args ...any) { LogPrintf(c.logger, "llm/openai ", format, args...) }
+func (c *openAIClient) logf(format string, args ...any) { logging.Logf("llm/openai ", format, args...) }
 
 func trimPreview(s string, n int) string {
 	if n <= 0 || len(s) <= n {
diff --git a/internal/llm/provider.go b/internal/llm/provider.go
index e83d1e2..a87d815 100644
--- a/internal/llm/provider.go
+++ b/internal/llm/provider.go
@@ -3,8 +3,7 @@ package llm
 import (
     "context"
     "errors"
-    "log"
-    "os"
+        "os"
 )
 
 // Message represents a chat-style prompt message.
@@ -40,10 +39,10 @@ func WithStop(stop ...string) RequestOption {
 
 // NewDefault returns the default provider using environment configuration.
 // Currently this is the OpenAI provider using OPENAI_API_KEY.
-func NewDefault(logger *log.Logger) (Client, error) {
+func NewDefault() (Client, error) {
     apiKey := os.Getenv("OPENAI_API_KEY")
     if apiKey == "" {
         return nil, errors.New("OPENAI_API_KEY is not set")
     }
-    return newOpenAIFromEnv(apiKey, logger), nil
+    return newOpenAIFromEnv(apiKey), nil
 }
diff --git a/internal/llm/logging.go b/internal/logging/logging.go
index a6a6e51..2e4bbc8 100644
--- a/internal/llm/logging.go
+++ b/internal/logging/logging.go
@@ -1,11 +1,11 @@
-package llm
+package logging
 
 import (
     "fmt"
     "log"
 )
 
-// ANSI color utilities shared across LLM providers.
+// ANSI color utilities shared across Hexai.
 const (
     AnsiBgBlack = "\x1b[40m"
     AnsiGrey    = "\x1b[90m"
@@ -18,13 +18,19 @@ const (
 // AnsiBase is the default style: black background + grey foreground.
 const AnsiBase = AnsiBgBlack + AnsiGrey
 
-// LogPrintf wraps a formatted message with a base style and prints with a prefix.
-func LogPrintf(logger *log.Logger, prefix, format string, args ...any) {
-    if logger == nil {
+// singleton logger used across the codebase
+var std *log.Logger
+
+// Bind sets the underlying standard logger to use for Logf.
+func Bind(l *log.Logger) { std = l }
+
+// Logf prints a formatted message with a module prefix and base ANSI style.
+func Logf(prefix, format string, args ...any) {
+    if std == nil {
         return
     }
     msg := fmt.Sprintf(format, args...)
-    logger.Print(AnsiBase + prefix + msg + AnsiReset)
+    std.Print(AnsiBase + prefix + msg + AnsiReset)
 }
 
 // Logging configuration for previews (shared)
@@ -34,9 +40,13 @@ var logPreviewLimit int // 0 means unlimited
 // request/response previews. Set to 0 for unlimited.
 func SetLogPreviewLimit(n int) { logPreviewLimit = n }
 
-func previewForLog(s string) string {
+// PreviewForLog returns the string truncated to the configured preview limit.
+func PreviewForLog(s string) string {
     if logPreviewLimit > 0 {
-        return trimPreview(s, logPreviewLimit)
+        if len(s) <= logPreviewLimit {
+            return s
+        }
+        return s[:logPreviewLimit] + "…"
     }
     return s
 }
diff --git a/internal/lsp/context.go b/internal/lsp/context.go
index 8b7ed67..8f345df 100644
--- a/internal/lsp/context.go
+++ b/internal/lsp/context.go
@@ -2,6 +2,7 @@ package lsp
 
 import (
     "strings"
+    "hexai/internal/logging"
 )
 
 // buildAdditionalContext builds extra context messages based on the configured mode.
@@ -33,9 +34,7 @@ func (s *Server) buildAdditionalContext(newFunc bool, uri string, pos Position)
 func (s *Server) windowContext(uri string, pos Position) string {
     d := s.getDocument(uri)
     if d == nil || len(d.lines) == 0 {
-        if s.logger != nil {
-            s.logger.Printf("context: window requested but document not open; skipping uri=%s", uri)
-        }
+        logging.Logf("lsp ", "context: window requested but document not open; skipping uri=%s", uri)
         return ""
     }
     n := len(d.lines)
@@ -55,9 +54,7 @@ func (s *Server) windowContext(uri string, pos Position) string {
 func (s *Server) fullFileContext(uri string) string {
     d := s.getDocument(uri)
     if d == nil {
-        if s.logger != nil {
-            s.logger.Printf("context: full-file requested but document not open; skipping uri=%s", uri)
-        }
+        logging.Logf("lsp ", "context: full-file requested but document not open; skipping uri=%s", uri)
         return ""
     }
     return truncateToApproxTokens(d.text, s.maxContextTokens)
diff --git a/internal/lsp/handlers.go b/internal/lsp/handlers.go
index 8a782c4..1e77141 100644
--- a/internal/lsp/handlers.go
+++ b/internal/lsp/handlers.go
@@ -1,14 +1,15 @@
 package lsp
 
 import (
-	"context"
-	"encoding/json"
-	"fmt"
-	"hexai/internal"
-	"hexai/internal/llm"
-	"os"
-	"strings"
-	"time"
+    "context"
+    "encoding/json"
+    "fmt"
+    "hexai/internal"
+    "hexai/internal/llm"
+    "hexai/internal/logging"
+    "os"
+    "strings"
+    "time"
 )
 
 func (s *Server) handle(req Request) {
@@ -52,7 +53,7 @@ func (s *Server) handleInitialize(req Request) {
 }
 
 func (s *Server) handleInitialized() {
-	s.logger.Println("client initialized")
+    logging.Logf("lsp ", "client initialized")
 }
 
 func (s *Server) handleShutdown(req Request) {
@@ -126,7 +127,7 @@ func (s *Server) buildDocString(p CompletionParams, above, current, below, funcC
 }
 
 func (s *Server) logCompletionContext(p CompletionParams, above, current, below, funcCtx string) {
-	s.logger.Printf("completion ctx uri=%s line=%d char=%d above=%q current=%q below=%q function=%q",
+	logging.Logf("lsp ", "completion ctx uri=%s line=%d char=%d above=%q current=%q below=%q function=%q",
 		p.TextDocument.URI, p.Position.Line, p.Position.Character, trimLen(above), trimLen(current), trimLen(below), trimLen(funcCtx))
 }
 
@@ -145,10 +146,10 @@ func (s *Server) tryLLMCompletion(p CompletionParams, above, current, below, fun
     }
 
     text, err := s.llmClient.Chat(ctx, messages, llm.WithMaxTokens(s.maxTokens), llm.WithTemperature(0.2))
-	if err != nil {
-		s.logger.Printf("llm completion error: %v", err)
-		return nil, false
-	}
+    if err != nil {
+        logging.Logf("lsp ", "llm completion error: %v", err)
+        return nil, false
+    }
 	cleaned := strings.TrimSpace(text)
 	if cleaned == "" {
 		return nil, false
diff --git a/internal/lsp/server.go b/internal/lsp/server.go
index 865d033..65d0b95 100644
--- a/internal/lsp/server.go
+++ b/internal/lsp/server.go
@@ -1,13 +1,14 @@
 package lsp
 
 import (
-	"bufio"
-	"encoding/json"
-	"hexai/internal/llm"
-	"io"
-	"log"
-	"sync"
-	"time"
+    "bufio"
+    "encoding/json"
+    "hexai/internal/llm"
+    "hexai/internal/logging"
+    "io"
+    "log"
+    "sync"
+    "time"
 )
 
 // Server implements a minimal LSP over stdio.
@@ -47,8 +48,8 @@ func NewServer(r io.Reader, w io.Writer, logger *log.Logger, logContext bool, ma
     s.windowLines = windowLines
     s.maxContextTokens = maxContextTokens
     s.noDiskIO = noDiskIO
-    if c, err := llm.NewDefault(logger); err != nil {
-        s.logger.Printf("llm disabled: %v", err)
+    if c, err := llm.NewDefault(); err != nil {
+        logging.Logf("lsp ", "llm disabled: %v", err)
     } else {
         s.llmClient = c
     }
@@ -66,9 +67,9 @@ func (s *Server) Run() error {
 		}
 		var req Request
 		if err := json.Unmarshal(body, &req); err != nil {
-			s.logger.Printf("invalid JSON: %v", err)
-			continue
-		}
+            logging.Logf("lsp ", "invalid JSON: %v", err)
+            continue
+        }
 		if req.Method == "" {
 			// A response from client; ignore
 			continue
diff --git a/internal/lsp/transport.go b/internal/lsp/transport.go
index 671d69b..dfdb5fc 100644
--- a/internal/lsp/transport.go
+++ b/internal/lsp/transport.go
@@ -1,12 +1,13 @@
 package lsp
 
 import (
-	"encoding/json"
-	"fmt"
-	"io"
-	"net/textproto"
-	"strconv"
-	"strings"
+    "encoding/json"
+    "fmt"
+    "hexai/internal/logging"
+    "io"
+    "net/textproto"
+    "strconv"
+    "strings"
 )
 
 func (s *Server) readMessage() ([]byte, error) {
@@ -47,17 +48,17 @@ func (s *Server) readMessage() ([]byte, error) {
 
 func (s *Server) writeMessage(v any) {
 	data, err := json.Marshal(v)
-	if err != nil {
-		s.logger.Printf("marshal error: %v", err)
-		return
-	}
+    if err != nil {
+        logging.Logf("lsp ", "marshal error: %v", err)
+        return
+    }
 	header := fmt.Sprintf("Content-Length: %d\r\n\r\n", len(data))
-	if _, err := io.WriteString(s.out, header); err != nil {
-		s.logger.Printf("write header error: %v", err)
-		return
-	}
-	if _, err := s.out.Write(data); err != nil {
-		s.logger.Printf("write body error: %v", err)
-		return
-	}
+    if _, err := io.WriteString(s.out, header); err != nil {
+        logging.Logf("lsp ", "write header error: %v", err)
+        return
+    }
+    if _, err := s.out.Write(data); err != nil {
+        logging.Logf("lsp ", "write body error: %v", err)
+        return
+    }
 }