Package restructure and API changes, several fixes

- More emphasis on `api` package. It now holds database model structs
  from `lmcli/models` (which is now gone) as well as the tool spec,
  call, and result types. `tools.Tool` is now `api.ToolSpec`.
  `api.ChatCompletionClient` was renamed to
  `api.ChatCompletionProvider`.

- Change ChatCompletion interface and implementations to no longer do
  automatic tool call recursion - they simply return a ToolCall message
  which the caller can decide what to do with (e.g. prompt for user
  confirmation before executing)

- `api.ChatCompletionProvider` functions have had their ReplyCallback
  parameter removed, as now they only return a single reply.

- Added a top-level `agent` package, moved the current built-in tools
  implementations under `agent/toolbox`. `tools.ExecuteToolCalls` is now
  `agent.ExecuteToolCalls`.

- Fixed request context handling in openai, google, ollama (use
  `NewRequestWithContext`), cleaned up request cancellation in TUI

- Fix tool call tui persistence bug (we were skipping message with empty
  content)

- Now handle tool calling from TUI layer

TODO:
- Prompt users before executing tool calls
- Automatically send tool results to the model (or make this toggleable)

pkg/tui/views/chat/update.go

@@ -4,7 +4,7 @@ import (
 	"strings"
 	"time"
 
-	models "git.mlow.ca/mlow/lmcli/pkg/lmcli/model"
+	"git.mlow.ca/mlow/lmcli/pkg/api"
 	"git.mlow.ca/mlow/lmcli/pkg/tui/shared"
 	tuiutil "git.mlow.ca/mlow/lmcli/pkg/tui/util"
 	"github.com/charmbracelet/bubbles/cursor"
@@ -21,15 +21,9 @@ func (m *Model) HandleResize(width, height int) {
 	}
 }
 
-func (m *Model) waitForResponse() tea.Cmd {
-	return func() tea.Msg {
-		return msgResponse(<-m.replyChan)
-	}
-}
-
 func (m *Model) waitForResponseChunk() tea.Cmd {
 	return func() tea.Msg {
-		return msgResponseChunk(<-m.replyChunkChan)
+		return msgChatResponseChunk(<-m.chatReplyChunks)
 	}
 }
 
@@ -48,7 +42,7 @@ func (m Model) Update(msg tea.Msg) (Model, tea.Cmd) {
 
 			if m.conversation.ShortName.String != m.Shared.Values.ConvShortname {
 				// clear existing messages if we're loading a new conversation
-				m.messages = []models.Message{}
+				m.messages = []api.Message{}
 				m.selectedMessage = 0
 			}
 		}
@@ -87,7 +81,7 @@ func (m Model) Update(msg tea.Msg) (Model, tea.Cmd) {
 		}
 		m.rebuildMessageCache()
 		m.updateContent()
-	case msgResponseChunk:
+	case msgChatResponseChunk:
 		cmds = append(cmds, m.waitForResponseChunk()) // wait for the next chunk
 
 		if msg.Content == "" {
@@ -100,8 +94,8 @@ func (m Model) Update(msg tea.Msg) (Model, tea.Cmd) {
 			m.setMessageContents(last, m.messages[last].Content+msg.Content)
 		} else {
 			// use chunk in new message
-			m.addMessage(models.Message{
-				Role:    models.MessageRoleAssistant,
+			m.addMessage(api.Message{
+				Role:    api.MessageRoleAssistant,
 				Content: msg.Content,
 			})
 		}
@@ -113,10 +107,10 @@ func (m Model) Update(msg tea.Msg) (Model, tea.Cmd) {
 
 		m.tokenCount += msg.TokenCount
 		m.elapsed = time.Now().Sub(m.startTime)
-	case msgResponse:
-		cmds = append(cmds, m.waitForResponse()) // wait for the next response
+	case msgChatResponse:
+		m.state = idle
 
-		reply := models.Message(msg)
+		reply := (*api.Message)(msg)
 		reply.Content = strings.TrimSpace(reply.Content)
 
 		last := len(m.messages) - 1
@@ -124,11 +118,18 @@ func (m Model) Update(msg tea.Msg) (Model, tea.Cmd) {
 			panic("Unexpected empty messages handling msgAssistantReply")
 		}
 
-		if reply.Role.IsAssistant() && m.messages[last].Role.IsAssistant() {
-			// this was a continuation, so replace the previous message with the completed reply
-			m.setMessage(last, reply)
+		if m.messages[last].Role.IsAssistant() {
+			// TODO: handle continuations gracefully - some models support them well, others fail horribly.
+			m.setMessage(last, *reply)
 		} else {
-			m.addMessage(reply)
+			m.addMessage(*reply)
+		}
+
+		switch reply.Role {
+		case api.MessageRoleToolCall:
+			// TODO: user confirmation before execution
+			// m.state = waitingForConfirmation
+			cmds = append(cmds, m.executeToolCalls(reply.ToolCalls))
 		}
 
 		if m.persistence {
@@ -140,17 +141,32 @@ func (m Model) Update(msg tea.Msg) (Model, tea.Cmd) {
 		}
 
 		m.updateContent()
-	case msgResponseEnd:
+	case msgChatResponseCanceled:
 		m.state = idle
-		last := len(m.messages) - 1
-		if last < 0 {
-			panic("Unexpected empty messages handling msgResponseEnd")
-		}
-		m.setMessageContents(last, strings.TrimSpace(m.messages[last].Content))
 		m.updateContent()
-	case msgResponseError:
+	case msgChatResponseError:
 		m.state = idle
 		m.Shared.Err = error(msg)
+		m.updateContent()
+	case msgToolResults:
+		last := len(m.messages) - 1
+		if last < 0 {
+			panic("Unexpected empty messages handling msgAssistantReply")
+		}
+
+		if m.messages[last].Role != api.MessageRoleToolCall {
+			panic("Previous message not a tool call, unexpected")
+		}
+
+		m.addMessage(api.Message{
+			Role:        api.MessageRoleToolResult,
+			ToolResults: api.ToolResults(msg),
+		})
+
+		if m.persistence {
+			cmds = append(cmds, m.persistConversation())
+		}
+
 		m.updateContent()
 	case msgConversationTitleGenerated:
 		title := string(msg)
@@ -167,7 +183,7 @@ func (m Model) Update(msg tea.Msg) (Model, tea.Cmd) {
 		m.conversation = msg.conversation
 		m.messages = msg.messages
 		if msg.isNew {
-			m.rootMessages = []models.Message{m.messages[0]}
+			m.rootMessages = []api.Message{m.messages[0]}
 		}
 		m.rebuildMessageCache()
 		m.updateContent()