Feature/cmd mode (#13)

Changes:
* Cmd Mode - it postprocess and attempts to execute llm output in bash
* Cleaning and improved error handling

.vscode/launch.json

@@ -78,6 +78,21 @@
       "env": {
         "NO_COLOR": "true"
       }
+    },
+    {
+      "name": "ChatGPT - Chat - Cmd",
+      "type": "go",
+      "request": "launch",
+      "program": "${workspaceFolder}",
+      "args": [
+        "-cm",
+        "gpt-4o",
+        "cmd",
+        "give me a command to show my current directory"
+      ],
+      "env": {
+        "NO_COLOR": "true"
+      }
     }
   ]
 }

go.mod

@@ -2,6 +2,6 @@ module github.com/baalimago/clai
 
 go 1.22
 
-require github.com/baalimago/go_away_boilerplate v1.3.10
+require github.com/baalimago/go_away_boilerplate v1.3.11
 
 require golang.org/x/net v0.24.0

go.sum

@@ -1,6 +1,4 @@
-github.com/baalimago/go_away_boilerplate v1.3.9 h1:oLtCiTNZAU2OM528QKxTsXyWXjTuK1j2YYVLXpyN7JQ=
-github.com/baalimago/go_away_boilerplate v1.3.9/go.mod h1:2O+zQ0Zm8vPD5SeccFFlgyf3AnYWQSHAut/ecPMmRdU=
-github.com/baalimago/go_away_boilerplate v1.3.10 h1:zID0+yZPimRZxw1XM8KcdBQ+IT9fbAqlA8pfYCe1Qrc=
-github.com/baalimago/go_away_boilerplate v1.3.10/go.mod h1:2O+zQ0Zm8vPD5SeccFFlgyf3AnYWQSHAut/ecPMmRdU=
+github.com/baalimago/go_away_boilerplate v1.3.11 h1:dM3Hckn55zSb6D9N03+TTSUxFJ6207GsCiS6n8vVZCs=
+github.com/baalimago/go_away_boilerplate v1.3.11/go.mod h1:2O+zQ0Zm8vPD5SeccFFlgyf3AnYWQSHAut/ecPMmRdU=
 golang.org/x/net v0.24.0 h1:1PcaxkF854Fu3+lvBIx5SYn9wRlBzzcnHZSiaFFAb0w=
 golang.org/x/net v0.24.0/go.mod h1:2Q7sJY5mzlzWjKtYUEXSlBWCdyaioyXzRB2RtU8KVE8=

internal/glob/glob.go

@@ -7,6 +7,7 @@ import (
 	"strings"
 
 	"github.com/baalimago/clai/internal/models"
+	"github.com/baalimago/clai/internal/utils"
 	"github.com/baalimago/go_away_boilerplate/pkg/ancli"
 	"github.com/baalimago/go_away_boilerplate/pkg/misc"
 )
@@ -64,11 +65,10 @@ func constructGlobMessages(globMessages []models.Message) []models.Message {
 }
 
 func parseGlob(glob string) ([]models.Message, error) {
-	home, err := os.UserHomeDir()
-	if err != nil && strings.Contains(glob, "~/") { // only fail if glob contains ~/ and home dir is not found
-		return nil, fmt.Errorf("failed to get home dir: %w", err)
+	glob, err := utils.ReplaceTildeWithHome(glob)
+	if err != nil {
+		return nil, fmt.Errorf("parseGlob, ReplaceTildeWithHome: %w", err)
 	}
-	glob = strings.Replace(glob, "~", home, 1)
 	files, err := filepath.Glob(glob)
 	ret := make([]models.Message, 0, len(files))
 	if err != nil {

internal/setup.go

@@ -15,6 +15,8 @@ import (
 	"github.com/baalimago/clai/internal/text"
 	"github.com/baalimago/clai/internal/utils"
 	"github.com/baalimago/go_away_boilerplate/pkg/ancli"
+	imagodebug "github.com/baalimago/go_away_boilerplate/pkg/debug"
+
 	"github.com/baalimago/go_away_boilerplate/pkg/misc"
 )
 
@@ -33,6 +35,7 @@ const (
 	PHOTO
 	VERSION
 	SETUP
+	CMD
 )
 
 var defaultFlags = Configurations{
@@ -66,6 +69,8 @@ func getModeFromArgs(cmd string) (Mode, error) {
 		return SETUP, nil
 	case "version", "v":
 		return VERSION, nil
+	case "cmd":
+		return CMD, nil
 	default:
 		return HELP, fmt.Errorf("unknown command: '%s'", os.Args[1])
 	}
@@ -81,13 +86,19 @@ func setupTextQuerier(mode Mode, confDir string, flagSet Configurations) (models
 	if mode == CHAT {
 		tConf.ChatMode = true
 	}
+
+	if mode == CMD {
+		tConf.CmdMode = true
+		tConf.SystemPrompt = tConf.CmdModePrompt
+	}
+
 	// At the moment, the configurations are based on the config file. But
 	// the configuration presecende is flags > file > default. So, we need
 	// to re-apply the flag overrides to the configuration
 	applyFlagOverridesForText(&tConf, flagSet, defaultFlags)
 
 	if misc.Truthy(os.Getenv("DEBUG")) {
-		ancli.PrintOK(fmt.Sprintf("config post flag override: %+v\n", tConf))
+		ancli.PrintOK(fmt.Sprintf("config post flag override: %+v\n", imagodebug.IndentedJsonFmt(tConf)))
 	}
 	args := flag.Args()
 	if mode == GLOB || flagSet.Glob != "" {
@@ -133,7 +144,7 @@ func Setup(usage string) (models.Querier, error) {
 	}
 
 	switch mode {
-	case CHAT, QUERY, GLOB:
+	case CHAT, QUERY, GLOB, CMD:
 		return setupTextQuerier(mode, confDir, flagSet)
 	case PHOTO:
 		pConf, err := utils.LoadConfigFromFile(confDir, "photoConfig.json", migrateOldPhotoConfig, &photo.DEFAULT)
@@ -153,7 +164,7 @@ func Setup(usage string) (models.Querier, error) {
 		}
 		pq, err := NewPhotoQuerier(pConf)
 		if misc.Truthy(os.Getenv("DEBUG")) {
-			ancli.PrintOK(fmt.Sprintf("photo querier: %+v\n", pq))
+			ancli.PrintOK(fmt.Sprintf("photo querier: %+v\n", imagodebug.IndentedJsonFmt(pq)))
 		}
 		if err != nil {
 			return nil, fmt.Errorf("failed to create photo querier: %v", err)

internal/text/conf.go

@@ -10,13 +10,15 @@ import (
 	"github.com/baalimago/clai/internal/reply"
 	"github.com/baalimago/clai/internal/utils"
 	"github.com/baalimago/go_away_boilerplate/pkg/ancli"
+	"github.com/baalimago/go_away_boilerplate/pkg/debug"
 	"github.com/baalimago/go_away_boilerplate/pkg/misc"
 )
 
 // Configurations used to setup the requirements of text models
 type Configurations struct {
 	Model          string      `json:"model"`
 	SystemPrompt   string      `json:"system-prompt"`
+	CmdModePrompt  string      `json:"cmd-mode-prompt"`
 	Raw            bool        `json:"raw"`
 	UseTools       bool        `json:"use-tools"`
 	TokenWarnLimit int         `json:"token-warn-limit"`
@@ -25,17 +27,19 @@ type Configurations struct {
 	Stream         bool        `json:"-"`
 	ReplyMode      bool        `json:"-"`
 	ChatMode       bool        `json:"-"`
+	CmdMode        bool        `json:"-"`
 	Glob           string      `json:"-"`
 	InitialPrompt  models.Chat `json:"-"`
 	// PostProccessedPrompt which has had it's strings replaced etc
 	PostProccessedPrompt string `json:"-"`
 }
 
 var DEFAULT = Configurations{
-	Model:        "gpt-4o",
-	SystemPrompt: "You are an assistant for a CLI interface. Answer concisely and informatively. Prefer markdown if possible.",
-	Raw:          false,
-	UseTools:     false,
+	Model:         "gpt-4o",
+	SystemPrompt:  "You are an assistant for a CLI tool. Answer concisely and informatively. Prefer markdown if possible.",
+	CmdModePrompt: "You are an assistant for a CLI tool aiding with cli tool suggestions. Write ONLY the command and nothing else.",
+	Raw:           false,
+	UseTools:      false,
 	// Aproximately $1 for the worst input rates as of 2024-05
 	TokenWarnLimit: 17000,
 }
@@ -45,6 +49,8 @@ func (c *Configurations) SetupPrompts(args []string) error {
 		ancli.PrintWarn("Using glob + reply modes together might yield strange results. The prevQuery will be appended after the glob messages.\n")
 	}
 
+	// Allways replace system prompt on cmd mode. This somewhat corrupts the chat since it always will
+	// be the command prompt. But it's better than not having it
 	if !c.ReplyMode {
 		c.InitialPrompt = models.Chat{
 			Messages: []models.Message{
@@ -69,6 +75,12 @@ func (c *Configurations) SetupPrompts(args []string) error {
 			return fmt.Errorf("failed to load previous query: %w", err)
 		}
 		c.InitialPrompt.Messages = append(c.InitialPrompt.Messages, iP.Messages...)
+
+		if c.CmdMode {
+			// Replace the initial message with the cmd prompt. This sort of
+			// destroys the history, but since the conversation might be long it's fine
+			c.InitialPrompt.Messages[0].Content = c.SystemPrompt
+		}
 	}
 
 	prompt, err := utils.Prompt(c.StdinReplace, args)
@@ -84,7 +96,7 @@ func (c *Configurations) SetupPrompts(args []string) error {
 	}
 
 	if misc.Truthy(os.Getenv("DEBUG")) {
-		ancli.PrintOK(fmt.Sprintf("InitialPrompt: %+v\n", c.InitialPrompt))
+		ancli.PrintOK(fmt.Sprintf("InitialPrompt: %v\n", debug.IndentedJsonFmt(c.InitialPrompt)))
 	}
 	c.PostProccessedPrompt = prompt
 	if c.InitialPrompt.ID == "" {

internal/text/generic/stream_completer.go

@@ -13,6 +13,7 @@ import (
 	"github.com/baalimago/clai/internal/models"
 	"github.com/baalimago/clai/internal/tools"
 	"github.com/baalimago/go_away_boilerplate/pkg/ancli"
+	"github.com/baalimago/go_away_boilerplate/pkg/debug"
 	"github.com/baalimago/go_away_boilerplate/pkg/misc"
 )
 
@@ -64,7 +65,7 @@ func (s *StreamCompleter) createRequest(ctx context.Context, chat models.Chat) (
 		reqData.ToolChoice = s.ToolChoice
 	}
 	if s.debug {
-		ancli.PrintOK(fmt.Sprintf("generic streamcompleter request: %+v\n", reqData))
+		ancli.PrintOK(fmt.Sprintf("generic streamcompleter request: %v\n", debug.IndentedJsonFmt(reqData)))
 	}
 	jsonData, err := json.Marshal(reqData)
 	if err != nil {

internal/text/querier.go

@@ -16,6 +16,7 @@ import (
 	"github.com/baalimago/clai/internal/tools"
 	"github.com/baalimago/clai/internal/utils"
 	"github.com/baalimago/go_away_boilerplate/pkg/ancli"
+	"github.com/baalimago/go_away_boilerplate/pkg/debug"
 )
 
 const (
@@ -38,32 +39,15 @@ type Querier[C models.StreamCompleter] struct {
 	hasPrinted      bool
 	Model           C
 	tokenWarnLimit  int
+	cmdMode         bool
 }
 
 // Query using the underlying model to stream completions and then print the output
 // from the model to stdout. Blocking operation.
 func (q *Querier[C]) Query(ctx context.Context) error {
-	amTokens := q.countTokens()
-	if q.tokenWarnLimit > 0 && amTokens > q.tokenWarnLimit {
-		ancli.PrintWarn(
-			fmt.Sprintf("You're about to send: ~%v tokens to the model, which may amount to: ~$%.3f (applying worst input rates as of 2024-05). This limit may be changed in: '%v'. Do you wish to continue? [yY]: ",
-				amTokens,
-				// Worst rates found at 2024-05 were gpt-4-32k at $60 per 1M tokens
-				float64(amTokens)*(float64(60)/float64(1000000)),
-				path.Join(q.configDir, "textConfig.json"),
-			))
-		var userInput string
-		reader := bufio.NewReader(os.Stdin)
-		userInput, err := reader.ReadString('\n')
-		if err != nil {
-			return fmt.Errorf("failed to read user input: %w", err)
-		}
-		switch userInput {
-		case "y\n", "Y\n":
-			// Continue on y or Y
-		default:
-			return errors.New("query canceled due to token amount check")
-		}
+	err := q.tokenLengthWarning()
+	if err != nil {
+		return fmt.Errorf("Querier.Query: %w", err)
 	}
 	completionsChan, err := q.Model.StreamCompletions(ctx, q.chat)
 	if err != nil {
@@ -106,6 +90,32 @@ func (q *Querier[C]) Query(ctx context.Context) error {
 	}
 }
 
+func (q *Querier[C]) tokenLengthWarning() error {
+	amTokens := q.countTokens()
+	if q.tokenWarnLimit > 0 && amTokens > q.tokenWarnLimit {
+		ancli.PrintWarn(
+			fmt.Sprintf("You're about to send: ~%v tokens to the model, which may amount to: ~$%.3f (applying worst input rates as of 2024-05). This limit may be changed in: '%v'. Do you wish to continue? [yY]: ",
+				amTokens,
+				// Worst rates found at 2024-05 were gpt-4-32k at $60 per 1M tokens
+				float64(amTokens)*(float64(60)/float64(1000000)),
+				path.Join(q.configDir, "textConfig.json"),
+			))
+		var userInput string
+		reader := bufio.NewReader(os.Stdin)
+		userInput, err := reader.ReadString('\n')
+		if err != nil {
+			return fmt.Errorf("failed to read user input: %w", err)
+		}
+		switch userInput {
+		case "y\n", "Y\n":
+			// Continue on y or Y
+		default:
+			return errors.New("Querier.tokenLengthWarning: query canceled due to token amount check")
+		}
+	}
+	return nil
+}
+
 // countTokens by simply counting the amount of strings which are delimited by whitespace
 // and multiply by some factor. This factor is somewhat arbritrary, and adjusted to be good enough
 // for all the different models
@@ -139,6 +149,23 @@ func (q *Querier[C]) postProcess() {
 		}
 	}
 
+	if q.debug {
+		ancli.PrintOK(fmt.Sprintf("Querier.postProcess:\n%v\n", debug.IndentedJsonFmt(q)))
+	}
+
+	// Cmd mode is a bit of a hack, it will handle all output
+	if q.cmdMode {
+		err := q.handleCmdMode()
+		if err != nil {
+			ancli.PrintErr(fmt.Sprintf("Querier.postProcess: %v\n", err))
+		}
+		return
+	}
+
+	q.postProcessOutput(newSysMsg)
+}
+
+func (q *Querier[C]) postProcessOutput(newSysMsg models.Message) {
 	// The token should already have been printed while streamed
 	if q.Raw {
 		return
@@ -163,13 +190,15 @@ func (q *Querier[C]) reset() {
 func (q *Querier[C]) TextQuery(ctx context.Context, chat models.Chat) (models.Chat, error) {
 	q.reset()
 	q.chat = chat
+	// Query will update the chat with the latest system message
 	err := q.Query(ctx)
 	if err != nil {
-		return models.Chat{}, fmt.Errorf("failed to query: %w", err)
+		return models.Chat{}, fmt.Errorf("TextQuery: %w", err)
 	}
 	if q.debug {
-		ancli.PrintOK(fmt.Sprintf("chat: %v", q.chat))
+		ancli.PrintOK(fmt.Sprintf("Querier.TextQuery:\n%v", debug.IndentedJsonFmt(q)))
 	}
+
 	return q.chat, nil
 }