chore: Bump version to 0.1.9

- Wire model_max_output_tokens config to Chat Completions API
- Config takes precedence over provider default for max_tokens

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

llmx-rs/Cargo.toml

@@ -43,7 +43,7 @@ members = [
 resolver = "2"
 
 [workspace.package]
-version = "0.1.2"
+version = "0.1.9"
 # Track the edition for all workspace crates in one place. Individual
 # crates can still override this value, but keeping it here means new
 # crates created with `cargo new -w ...` automatically inherit the 2024

llmx-rs/app-server/tests/common/mcp_process.rs

@@ -138,7 +138,7 @@ impl McpProcess {
             client_info: ClientInfo {
                 name: "llmx-app-server-tests".to_string(),
                 title: None,
-                version: "0.1.2".to_string(),
+                version: "0.1.7".to_string(),
             },
         })?);
         let req_id = self.send_request("initialize", params).await?;

llmx-rs/app-server/tests/suite/user_agent.rs

@@ -26,7 +26,7 @@ async fn get_user_agent_returns_current_llmx_user_agent() -> Result<()> {
 
     let os_info = os_info::get();
     let user_agent = format!(
-        "llmx_cli_rs/0.1.2 ({} {}; {}) {} (llmx-app-server-tests; 0.1.2)",
+        "llmx_cli_rs/0.1.7 ({} {}; {}) {} (llmx-app-server-tests; 0.1.7)",
         os_info.os_type(),
         os_info.version(),
         os_info.architecture().unwrap_or("unknown"),

llmx-rs/apply-patch/src/parser.rs

@@ -31,6 +31,7 @@ use thiserror::Error;
 const BEGIN_PATCH_MARKER: &str = "*** Begin Patch";
 const END_PATCH_MARKER: &str = "*** End Patch";
 const ADD_FILE_MARKER: &str = "*** Add File: ";
+const CREATE_FILE_MARKER: &str = "*** Create File: "; // Alias for Add File
 const DELETE_FILE_MARKER: &str = "*** Delete File: ";
 const UPDATE_FILE_MARKER: &str = "*** Update File: ";
 const MOVE_TO_MARKER: &str = "*** Move to: ";
@@ -245,8 +246,8 @@ fn check_start_and_end_lines_strict(
 fn parse_one_hunk(lines: &[&str], line_number: usize) -> Result<(Hunk, usize), ParseError> {
     // Be tolerant of case mismatches and extra padding around marker strings.
     let first_line = lines[0].trim();
-    if let Some(path) = first_line.strip_prefix(ADD_FILE_MARKER) {
+    if let Some(path) = first_line.strip_prefix(ADD_FILE_MARKER).or_else(|| first_line.strip_prefix(CREATE_FILE_MARKER)) {
-        // Add File
+        // Add File (also accepts Create File as alias)
         let mut contents = String::new();
         let mut parsed_lines = 1;
         for add_line in &lines[1..] {
@@ -331,7 +332,7 @@ fn parse_one_hunk(lines: &[&str], line_number: usize) -> Result<(Hunk, usize), P
 
     Err(InvalidHunkError {
         message: format!(
-            "'{first_line}' is not a valid hunk header. Valid hunk headers: '*** Add File: {{path}}', '*** Delete File: {{path}}', '*** Update File: {{path}}'"
+            "'{first_line}' is not a valid hunk header. Valid hunk headers: '*** Add File: {{path}}', '*** Create File: {{path}}', '*** Delete File: {{path}}', '*** Update File: {{path}}'"
         ),
         line_number,
     })

llmx-rs/core/src/chat_completions.rs

@@ -45,6 +45,7 @@ pub(crate) async fn stream_chat_completions(
     provider: &ModelProviderInfo,
     otel_event_manager: &OtelEventManager,
     session_source: &SessionSource,
+    model_max_output_tokens: Option<i64>,
 ) -> Result<ResponseStream> {
     if prompt.output_schema.is_some() {
         return Err(LlmxErr::UnsupportedOperation(
@@ -56,7 +57,12 @@ pub(crate) async fn stream_chat_completions(
     let mut messages = Vec::<serde_json::Value>::new();
 
     let full_instructions = prompt.get_full_instructions(model_family);
-    messages.push(json!({"role": "system", "content": full_instructions}));
+    // Add cache_control to system instructions for Anthropic prompt caching
+    messages.push(json!({
+        "role": "system",
+        "content": full_instructions,
+        "cache_control": {"type": "ephemeral"}
+    }));
 
     let input = prompt.get_formatted_input();
 
@@ -161,7 +167,65 @@ pub(crate) async fn stream_chat_completions(
     // aggregated assistant message was recorded alongside an earlier partial).
     let mut last_assistant_text: Option<String> = None;
 
+    // Build a map of which call_ids have outputs
+    // We'll use this to ensure we never send a FunctionCall without its corresponding output
+    let mut call_ids_with_outputs: std::collections::HashSet<String> = std::collections::HashSet::new();
+
+    // First pass: collect all call_ids that have outputs
+    for item in input.iter() {
+        if let ResponseItem::FunctionCallOutput { call_id, .. } = item {
+            call_ids_with_outputs.insert(call_id.clone());
+        }
+    }
+
+    debug!("=== Chat Completions Request Debug ===");
+    debug!("Input items count: {}", input.len());
+    debug!("Call IDs with outputs: {:?}", call_ids_with_outputs);
+
+    // Second pass: find the first FunctionCall that doesn't have an output
+    let mut cutoff_at_idx: Option<usize> = None;
     for (idx, item) in input.iter().enumerate() {
+        if let ResponseItem::FunctionCall { call_id, name, .. } = item {
+            if !call_ids_with_outputs.contains(call_id) {
+                debug!("Found unanswered function call '{}' (call_id: {}) at index {}", name, call_id, idx);
+                cutoff_at_idx = Some(idx);
+                break;
+            }
+        }
+    }
+
+    if let Some(cutoff) = cutoff_at_idx {
+        debug!("Cutting off at index {} to avoid orphaned tool calls", cutoff);
+    } else {
+        debug!("No unanswered function calls found, processing all items");
+    }
+
+    // Track whether the MOST RECENT FunctionCall with each call_id was skipped
+    // This allows the same call_id to be retried - we only skip outputs for the specific skipped calls
+    let mut call_id_skip_state: std::collections::HashMap<String, bool> = std::collections::HashMap::new();
+
+    for (idx, item) in input.iter().enumerate() {
+        // Stop processing if we've reached an unanswered function call
+        if let Some(cutoff) = cutoff_at_idx {
+            if idx >= cutoff {
+                debug!("Stopping at index {} due to unanswered function call", idx);
+                break;
+            }
+        }
+
+        debug!("Processing item {} of type: {}", idx, match item {
+            ResponseItem::Message { role, .. } => format!("Message(role={})", role),
+            ResponseItem::FunctionCall { name, call_id, .. } => format!("FunctionCall(name={}, call_id={})", name, call_id),
+            ResponseItem::FunctionCallOutput { call_id, .. } => format!("FunctionCallOutput(call_id={})", call_id),
+            ResponseItem::LocalShellCall { .. } => "LocalShellCall".to_string(),
+            ResponseItem::CustomToolCall { .. } => "CustomToolCall".to_string(),
+            ResponseItem::CustomToolCallOutput { .. } => "CustomToolCallOutput".to_string(),
+            ResponseItem::Reasoning { .. } => "Reasoning".to_string(),
+            ResponseItem::WebSearchCall { .. } => "WebSearchCall".to_string(),
+            ResponseItem::GhostSnapshot { .. } => "GhostSnapshot".to_string(),
+            ResponseItem::Other => "Other".to_string(),
+        });
+
         match item {
             ResponseItem::Message { role, content, .. } => {
                 // Build content either as a plain string (typical for assistant text)
@@ -175,8 +239,11 @@ pub(crate) async fn stream_chat_completions(
                         ContentItem::InputText { text: t }
                         | ContentItem::OutputText { text: t } => {
                             text.push_str(t);
+                            // Only add text content blocks that are non-empty
+                            if !t.trim().is_empty() {
                                 items.push(json!({"type":"text","text": t}));
                             }
+                        }
                         ContentItem::InputImage { image_url } => {
                             saw_image = true;
                             items.push(json!({"type":"image_url","image_url": {"url": image_url}}));
@@ -184,6 +251,11 @@ pub(crate) async fn stream_chat_completions(
                     }
                 }
 
+                // Skip messages with empty or whitespace-only text content (unless they contain images)
+                if text.trim().is_empty() && !saw_image {
+                    continue;
+                }
+
                 // Skip exact-duplicate assistant messages.
                 if role == "assistant" {
                     if let Some(prev) = &last_assistant_text
@@ -219,6 +291,18 @@ pub(crate) async fn stream_chat_completions(
                 call_id,
                 ..
             } => {
+                // Validate that arguments is valid JSON before sending to API
+                // If invalid, skip this function call to avoid API errors
+                if serde_json::from_str::<serde_json::Value>(arguments).is_err() {
+                    debug!("Skipping malformed function call with invalid JSON arguments: {}", arguments);
+                    // Mark this call_id's most recent state as skipped
+                    call_id_skip_state.insert(call_id.clone(), true);
+                    continue;
+                }
+
+                // Mark this call_id's most recent state as NOT skipped (valid call)
+                call_id_skip_state.insert(call_id.clone(), false);
+
                 let mut msg = json!({
                     "role": "assistant",
                     "content": null,
@@ -263,6 +347,12 @@ pub(crate) async fn stream_chat_completions(
                 messages.push(msg);
             }
             ResponseItem::FunctionCallOutput { call_id, output } => {
+                // Skip outputs only if the MOST RECENT FunctionCall with this call_id was skipped
+                if call_id_skip_state.get(call_id) == Some(&true) {
+                    debug!("Skipping function call output for most recent skipped call_id: {}", call_id);
+                    continue;
+                }
+
                 // Prefer structured content items when available (e.g., images)
                 // otherwise fall back to the legacy plain-string content.
                 let content_value = if let Some(items) = &output.content_items {
@@ -328,14 +418,41 @@ pub(crate) async fn stream_chat_completions(
         }
     }
 
+    debug!("Built {} messages for API request", messages.len());
+
+    // Add cache_control to conversation history for Anthropic prompt caching
+    // Add it to a message that's at least 3 messages before the end (stable history)
+    // This caches the earlier conversation while keeping recent turns uncached
+    if messages.len() > 4 {
+        let cache_idx = messages.len().saturating_sub(4);
+        if let Some(msg) = messages.get_mut(cache_idx) {
+            if let Some(obj) = msg.as_object_mut() {
+                obj.insert("cache_control".to_string(), json!({"type": "ephemeral"}));
+                debug!("Added cache_control to message at index {} (conversation history)", cache_idx);
+            }
+        }
+    }
+
+    debug!("=== End Chat Completions Request Debug ===");
+
     let tools_json = create_tools_json_for_chat_completions_api(&prompt.tools)?;
-    let payload = json!({
+    let mut payload = json!({
         "model": model_family.slug,
         "messages": messages,
         "stream": true,
         "tools": tools_json,
     });
 
+    // Add max_tokens - required by Anthropic Messages API
+    // Priority: config model_max_output_tokens > provider max_tokens > default 20480
+    let max_tokens = model_max_output_tokens
+        .or(provider.max_tokens)
+        .unwrap_or(20480);
+    if let Some(obj) = payload.as_object_mut() {
+        obj.insert("max_tokens".to_string(), json!(max_tokens));
+    }
+    debug!("Using max_tokens: {}", max_tokens);
+
     debug!(
         "POST to {}: {}",
         provider.get_full_url(&None),
@@ -496,7 +613,9 @@ async fn process_chat_sse<S>(
 ) where
     S: Stream<Item = Result<Bytes>> + Unpin,
 {
+    debug!("process_chat_sse started, idle_timeout={:?}", idle_timeout);
     let mut stream = stream.eventsource();
+    debug!("SSE stream initialized, waiting for first event");
 
     // State to accumulate a function call across streaming chunks.
     // OpenAI may split the `arguments` string over multiple `delta` events
@@ -531,7 +650,14 @@ async fn process_chat_sse<S>(
                 return;
             }
             Ok(None) => {
-                // Stream closed gracefully – emit Completed with dummy id.
+                // Stream closed gracefully – emit any pending items first, then Completed
+                debug!("Stream closed gracefully (Ok(None)), emitting pending items");
+                if let Some(item) = assistant_item.take() {
+                    let _ = tx_event.send(Ok(ResponseEvent::OutputItemDone(item))).await;
+                }
+                if let Some(item) = reasoning_item.take() {
+                    let _ = tx_event.send(Ok(ResponseEvent::OutputItemDone(item))).await;
+                }
                 let _ = tx_event
                     .send(Ok(ResponseEvent::Completed {
                         response_id: String::new(),
@@ -727,6 +853,7 @@ async fn process_chat_sse<S>(
 
             // Emit end-of-turn when finish_reason signals completion.
             if let Some(finish_reason) = choice.get("finish_reason").and_then(|v| v.as_str()) {
+                debug!("Received finish_reason: {}", finish_reason);
                 match finish_reason {
                     "tool_calls" if fn_call_state.active => {
                         // First, flush the terminal raw reasoning so UIs can finalize
@@ -745,27 +872,46 @@ async fn process_chat_sse<S>(
 
                         let _ = tx_event.send(Ok(ResponseEvent::OutputItemDone(item))).await;
                     }
-                    "stop" => {
+                    "stop" | "length" => {
-                        // Regular turn without tool-call. Emit the final assistant message
+                        // Regular turn without tool-call, or hit max_tokens limit.
-                        // as a single OutputItemDone so non-delta consumers see the result.
+                        debug!("Processing finish_reason={}, assistant_item.is_some()={}, reasoning_item.is_some()={}",
+                            finish_reason, assistant_item.is_some(), reasoning_item.is_some());
+                        // Emit the final assistant message as a single OutputItemDone so non-delta consumers see the result.
+                        if let Some(item) = assistant_item.take() {
+                            debug!("Emitting assistant_item: {:?}", item);
+                            let _ = tx_event.send(Ok(ResponseEvent::OutputItemDone(item))).await;
+                        } else {
+                            debug!("No assistant_item to emit");
+                        }
+                        // Also emit a terminal Reasoning item so UIs can finalize raw reasoning.
+                        if let Some(item) = reasoning_item.take() {
+                            debug!("Emitting reasoning_item");
+                            let _ = tx_event.send(Ok(ResponseEvent::OutputItemDone(item))).await;
+                        } else {
+                            debug!("No reasoning_item to emit");
+                        }
+                    }
+                    _ => {
+                        // Unknown finish_reason - still emit pending items to avoid hanging
+                        debug!("Unknown finish_reason: {}, emitting pending items", finish_reason);
                         if let Some(item) = assistant_item.take() {
                             let _ = tx_event.send(Ok(ResponseEvent::OutputItemDone(item))).await;
                         }
-                        // Also emit a terminal Reasoning item so UIs can finalize raw reasoning.
                         if let Some(item) = reasoning_item.take() {
                             let _ = tx_event.send(Ok(ResponseEvent::OutputItemDone(item))).await;
                         }
                     }
-                    _ => {}
                 }
 
                 // Emit Completed regardless of reason so the agent can advance.
+                debug!("Sending Completed event after finish_reason={}", finish_reason);
                 let _ = tx_event
                     .send(Ok(ResponseEvent::Completed {
                         response_id: String::new(),
                         token_usage: token_usage.clone(),
                     }))
                     .await;
+                debug!("Completed event sent, returning from SSE processor");
 
                 // Prepare for potential next turn (should not happen in same stream).
                 // fn_call_state = FunctionCallState::default();
@@ -774,6 +920,22 @@ async fn process_chat_sse<S>(
             }
         }
     }
+
+    // Stream ended without finish_reason - this can happen when the stream closes abruptly
+    debug!("Stream ended without finish_reason, emitting final items and Completed event");
+    if let Some(item) = assistant_item.take() {
+        let _ = tx_event.send(Ok(ResponseEvent::OutputItemDone(item))).await;
+    }
+    if let Some(item) = reasoning_item.take() {
+        let _ = tx_event.send(Ok(ResponseEvent::OutputItemDone(item))).await;
+    }
+    // Send Completed event so llmx knows the turn is done
+    let _ = tx_event
+        .send(Ok(ResponseEvent::Completed {
+            response_id: String::new(),
+            token_usage: token_usage.clone(),
+        }))
+        .await;
 }
 
 /// Optional client-side aggregation helper

llmx-rs/core/src/client.rs

@@ -152,6 +152,7 @@ impl ModelClient {
                     &self.provider,
                     &self.otel_event_manager,
                     &self.session_source,
+                    self.config.model_max_output_tokens,
                 )
                 .await?;
 
@@ -1123,6 +1124,7 @@ mod tests {
             request_max_retries: Some(0),
             stream_max_retries: Some(0),
             stream_idle_timeout_ms: Some(1000),
+        max_tokens: None,
             requires_openai_auth: false,
         };
 
@@ -1187,6 +1189,7 @@ mod tests {
             request_max_retries: Some(0),
             stream_max_retries: Some(0),
             stream_idle_timeout_ms: Some(1000),
+        max_tokens: None,
             requires_openai_auth: false,
         };
 
@@ -1224,6 +1227,7 @@ mod tests {
             request_max_retries: Some(0),
             stream_max_retries: Some(0),
             stream_idle_timeout_ms: Some(1000),
+        max_tokens: None,
             requires_openai_auth: false,
         };
 
@@ -1263,6 +1267,7 @@ mod tests {
             request_max_retries: Some(0),
             stream_max_retries: Some(0),
             stream_idle_timeout_ms: Some(1000),
+        max_tokens: None,
             requires_openai_auth: false,
         };
 
@@ -1298,6 +1303,7 @@ mod tests {
             request_max_retries: Some(0),
             stream_max_retries: Some(0),
             stream_idle_timeout_ms: Some(1000),
+        max_tokens: None,
             requires_openai_auth: false,
         };
 
@@ -1333,6 +1339,7 @@ mod tests {
             request_max_retries: Some(0),
             stream_max_retries: Some(0),
             stream_idle_timeout_ms: Some(1000),
+        max_tokens: None,
             requires_openai_auth: false,
         };
 
@@ -1437,6 +1444,7 @@ mod tests {
                 request_max_retries: Some(0),
                 stream_max_retries: Some(0),
                 stream_idle_timeout_ms: Some(1000),
+        max_tokens: None,
                 requires_openai_auth: false,
             };
 

llmx-rs/core/src/config/mod.rs

@@ -973,6 +973,8 @@ impl Config {
 
         let mut model_providers = built_in_model_providers();
         // Merge user-defined providers into the built-in list.
+        // Note: This uses or_insert() so built-in providers take precedence.
+        // For custom max_tokens, use model_max_output_tokens config instead.
         for (key, provider) in cfg.model_providers.into_iter() {
             model_providers.entry(key).or_insert(provider);
         }
@@ -2809,6 +2811,7 @@ model_verbosity = "high"
             request_max_retries: Some(4),
             stream_max_retries: Some(10),
             stream_idle_timeout_ms: Some(300_000),
+        max_tokens: None,
             requires_openai_auth: false,
         };
         let model_provider_map = {

llmx-rs/core/src/event_mapping.rs

@@ -54,7 +54,7 @@ fn parse_user_message(message: &[ContentItem]) -> Option<UserMessageItem> {
     Some(UserMessageItem::new(&content))
 }
 
-fn parse_agent_message(id: Option<&String>, message: &[ContentItem]) -> AgentMessageItem {
+fn parse_agent_message(id: Option<&String>, message: &[ContentItem]) -> Option<AgentMessageItem> {
     let mut content: Vec<AgentMessageContent> = Vec::new();
     for content_item in message.iter() {
         match content_item {
@@ -69,18 +69,23 @@ fn parse_agent_message(id: Option<&String>, message: &[ContentItem]) -> AgentMes
             }
         }
     }
+
+    // If the message has no content, return None to signal turn completion
+    // This happens when the API ends a turn with an empty assistant message (e.g., after tool calls)
+    if content.is_empty() {
+        return None;
+    }
+
     let id = id.cloned().unwrap_or_else(|| Uuid::new_v4().to_string());
-    AgentMessageItem { id, content }
+    Some(AgentMessageItem { id, content })
 }
 
 pub fn parse_turn_item(item: &ResponseItem) -> Option<TurnItem> {
     match item {
         ResponseItem::Message { role, content, id } => match role.as_str() {
             "user" => parse_user_message(content).map(TurnItem::UserMessage),
-            "assistant" => Some(TurnItem::AgentMessage(parse_agent_message(
+            "assistant" => parse_agent_message(id.as_ref(), content)
-                id.as_ref(),
+                .map(TurnItem::AgentMessage),
-                content,
-            ))),
             "system" => None,
             _ => None,
         },

llmx-rs/core/src/model_provider_info.rs

@@ -87,6 +87,10 @@ pub struct ModelProviderInfo {
     /// the connection as lost.
     pub stream_idle_timeout_ms: Option<u64>,
 
+    /// Maximum number of tokens to generate in the response. If not specified, defaults to 8192.
+    /// This is required by some providers (e.g., Anthropic via LiteLLM).
+    pub max_tokens: Option<i64>,
+
     /// Does this provider require an OpenAI API Key or ChatGPT login token? If true,
     /// user is presented with login screen on first run, and login preference and token/key
     /// are stored in auth.json. If false (which is the default), login screen is skipped,
@@ -290,6 +294,7 @@ pub fn built_in_model_providers() -> HashMap<String, ModelProviderInfo> {
                 request_max_retries: None,
                 stream_max_retries: None,
                 stream_idle_timeout_ms: None,
+            max_tokens: None,
                 requires_openai_auth: false,
             },
         ),
@@ -330,6 +335,7 @@ pub fn built_in_model_providers() -> HashMap<String, ModelProviderInfo> {
                 request_max_retries: None,
                 stream_max_retries: None,
                 stream_idle_timeout_ms: None,
+            max_tokens: None,
                 requires_openai_auth: true,
             },
         ),
@@ -375,6 +381,7 @@ pub fn create_oss_provider_with_base_url(base_url: &str) -> ModelProviderInfo {
         request_max_retries: None,
         stream_max_retries: None,
         stream_idle_timeout_ms: None,
+            max_tokens: None,
         requires_openai_auth: false,
     }
 }
@@ -415,6 +422,7 @@ base_url = "http://localhost:11434/v1"
             request_max_retries: None,
             stream_max_retries: None,
             stream_idle_timeout_ms: None,
+            max_tokens: None,
             requires_openai_auth: false,
         };
 
@@ -445,6 +453,7 @@ query_params = { api-version = "2025-04-01-preview" }
             request_max_retries: None,
             stream_max_retries: None,
             stream_idle_timeout_ms: None,
+            max_tokens: None,
             requires_openai_auth: false,
         };
 
@@ -478,6 +487,7 @@ env_http_headers = { "X-Example-Env-Header" = "EXAMPLE_ENV_VAR" }
             request_max_retries: None,
             stream_max_retries: None,
             stream_idle_timeout_ms: None,
+            max_tokens: None,
             requires_openai_auth: false,
         };
 
@@ -501,6 +511,7 @@ env_http_headers = { "X-Example-Env-Header" = "EXAMPLE_ENV_VAR" }
                 request_max_retries: None,
                 stream_max_retries: None,
                 stream_idle_timeout_ms: None,
+            max_tokens: None,
                 requires_openai_auth: false,
             }
         }
@@ -534,6 +545,7 @@ env_http_headers = { "X-Example-Env-Header" = "EXAMPLE_ENV_VAR" }
             request_max_retries: None,
             stream_max_retries: None,
             stream_idle_timeout_ms: None,
+            max_tokens: None,
             requires_openai_auth: false,
         };
         assert!(named_provider.is_azure_responses_endpoint());

llmx-rs/core/src/tools/spec.rs

@@ -693,7 +693,7 @@ pub(crate) fn create_tools_json_for_chat_completions_api(
     // We start with the JSON for the Responses API and than rewrite it to match
     // the chat completions tool call format.
     let responses_api_tools_json = create_tools_json_for_responses_api(tools)?;
-    let tools_json = responses_api_tools_json
+    let mut tools_json = responses_api_tools_json
         .into_iter()
         .filter_map(|mut tool| {
             if tool.get("type") != Some(&serde_json::Value::String("function".to_string())) {
@@ -712,6 +712,14 @@ pub(crate) fn create_tools_json_for_chat_completions_api(
             }
         })
         .collect::<Vec<serde_json::Value>>();
+
+    // Add cache_control to the last tool to enable Anthropic prompt caching
+    if let Some(last_tool) = tools_json.last_mut() {
+        if let Some(obj) = last_tool.as_object_mut() {
+            obj.insert("cache_control".to_string(), json!({"type": "ephemeral"}));
+        }
+    }
+
     Ok(tools_json)
 }
 

llmx-rs/core/tests/chat_completions_payload.rs

@@ -58,6 +58,7 @@ async fn run_request(input: Vec<ResponseItem>) -> Value {
         request_max_retries: Some(0),
         stream_max_retries: Some(0),
         stream_idle_timeout_ms: Some(5_000),
+        max_tokens: None,
         requires_openai_auth: false,
     };
 

llmx-rs/core/tests/chat_completions_sse.rs

@@ -58,6 +58,7 @@ async fn run_stream_with_bytes(sse_body: &[u8]) -> Vec<ResponseEvent> {
         request_max_retries: Some(0),
         stream_max_retries: Some(0),
         stream_idle_timeout_ms: Some(5_000),
+        max_tokens: None,
         requires_openai_auth: false,
     };
 

llmx-rs/core/tests/responses_headers.rs

@@ -47,6 +47,7 @@ async fn responses_stream_includes_subagent_header_on_review() {
         request_max_retries: Some(0),
         stream_max_retries: Some(0),
         stream_idle_timeout_ms: Some(5_000),
+        max_tokens: None,
         requires_openai_auth: false,
     };
 
@@ -135,6 +136,7 @@ async fn responses_stream_includes_subagent_header_on_other() {
         request_max_retries: Some(0),
         stream_max_retries: Some(0),
         stream_idle_timeout_ms: Some(5_000),
+        max_tokens: None,
         requires_openai_auth: false,
     };
 

llmx-rs/core/tests/suite/client.rs

@@ -712,6 +712,7 @@ async fn azure_responses_request_includes_store_and_reasoning_ids() {
         request_max_retries: Some(0),
         stream_max_retries: Some(0),
         stream_idle_timeout_ms: Some(5_000),
+        max_tokens: None,
         requires_openai_auth: false,
     };
 
@@ -1195,6 +1196,7 @@ async fn azure_overrides_assign_properties_used_for_responses_url() {
         request_max_retries: None,
         stream_max_retries: None,
         stream_idle_timeout_ms: None,
+        max_tokens: None,
         requires_openai_auth: false,
     };
 
@@ -1272,6 +1274,7 @@ async fn env_var_overrides_loaded_auth() {
         request_max_retries: None,
         stream_max_retries: None,
         stream_idle_timeout_ms: None,
+        max_tokens: None,
         requires_openai_auth: false,
     };
 

llmx-rs/core/tests/suite/stream_error_allows_next_turn.rs

@@ -72,6 +72,7 @@ async fn continue_after_stream_error() {
         request_max_retries: Some(1),
         stream_max_retries: Some(1),
         stream_idle_timeout_ms: Some(2_000),
+        max_tokens: None,
         requires_openai_auth: false,
     };
 

llmx-rs/core/tests/suite/stream_no_completed.rs

@@ -80,6 +80,7 @@ async fn retries_on_early_close() {
         request_max_retries: Some(0),
         stream_max_retries: Some(1),
         stream_idle_timeout_ms: Some(2000),
+        max_tokens: None,
         requires_openai_auth: false,
     };
 

llmx-rs/mcp-server/tests/common/mcp_process.rs

@@ -144,7 +144,7 @@ impl McpProcess {
         let initialized = self.read_jsonrpc_message().await?;
         let os_info = os_info::get();
         let user_agent = format!(
-            "llmx_cli_rs/0.1.2 ({} {}; {}) {} (elicitation test; 0.0.0)",
+            "llmx_cli_rs/0.1.7 ({} {}; {}) {} (elicitation test; 0.0.0)",
             os_info.os_type(),
             os_info.version(),
             os_info.architecture().unwrap_or("unknown"),
@@ -163,7 +163,7 @@ impl McpProcess {
                     "serverInfo": {
                         "name": "llmx-mcp-server",
                         "title": "LLMX",
-                        "version": "0.1.2",
+                        "version": "0.1.7",
                         "user_agent": user_agent
                     },
                     "protocolVersion": mcp_types::MCP_SCHEMA_VERSION

llmx-rs/tui/src/status/snapshots/llmx_tui__status__tests__status_snapshot_includes_monthly_limit.snap

@@ -5,7 +5,7 @@ expression: sanitized
 /status
 
 ╭───────────────────────────────────────────────────────────────────────────╮
-│  >_ LLMX (v0.1.2)                                                         │
+│  >_ LLMX (v0.1.7)                                                         │
 │                                                                           │
 │ Visit https://chatgpt.com/llmx/settings/usage for up-to-date              │
 │ information on rate limits and credits                                    │

llmx-rs/tui/src/status/snapshots/llmx_tui__status__tests__status_snapshot_includes_reasoning_details.snap

@@ -5,7 +5,7 @@ expression: sanitized
 /status
 
 ╭─────────────────────────────────────────────────────────────────╮
-│  >_ LLMX (v0.1.2)                                               │
+│  >_ LLMX (v0.1.7)                                               │
 │                                                                 │
 │ Visit https://chatgpt.com/llmx/settings/usage for up-to-date    │
 │ information on rate limits and credits                          │

llmx-rs/tui/src/status/snapshots/llmx_tui__status__tests__status_snapshot_shows_empty_limits_message.snap

@@ -5,7 +5,7 @@ expression: sanitized
 /status
 
 ╭──────────────────────────────────────────────────────────────╮
-│  >_ LLMX (v0.1.2)                                            │
+│  >_ LLMX (v0.1.7)                                            │
 │                                                              │
 │ Visit https://chatgpt.com/llmx/settings/usage for up-to-date │
 │ information on rate limits and credits                       │

llmx-rs/tui/src/status/snapshots/llmx_tui__status__tests__status_snapshot_shows_missing_limits_message.snap

@@ -5,7 +5,7 @@ expression: sanitized
 /status
 
 ╭──────────────────────────────────────────────────────────────╮
-│  >_ LLMX (v0.1.2)                                            │
+│  >_ LLMX (v0.1.7)                                            │
 │                                                              │
 │ Visit https://chatgpt.com/llmx/settings/usage for up-to-date │
 │ information on rate limits and credits                       │

llmx-rs/tui/src/status/snapshots/llmx_tui__status__tests__status_snapshot_shows_stale_limits_message.snap

@@ -5,7 +5,7 @@ expression: sanitized
 /status
 
 ╭───────────────────────────────────────────────────────────────────╮
-│  >_ LLMX (v0.1.2)                                                 │
+│  >_ LLMX (v0.1.7)                                                 │
 │                                                                   │
 │ Visit https://chatgpt.com/llmx/settings/usage for up-to-date      │
 │ information on rate limits and credits                            │

llmx-rs/tui/src/status/snapshots/llmx_tui__status__tests__status_snapshot_truncates_in_narrow_terminal.snap

@@ -5,7 +5,7 @@ expression: sanitized
 /status
 
 ╭────────────────────────────────────────────╮
-│  >_ LLMX (v0.1.2)                          │
+│  >_ LLMX (v0.1.7)                          │
 │                                            │
 │ Visit https://chatgpt.com/llmx/settings/   │
 │ usage for up-to-date                       │