lxy/DevFlow
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

优化: token分项显示(in/cache/out/reasoning)+ 详情面板 + base前置

Browse Source 
token分项(各计费不同,不显 total):df-ai 解析 provider cache/reasoning(openai_compat prompt_cache_hit/miss/reasoning_tokens + anthropic cache_read/creation)+ TokenUsage 加字段(全构造点)+ AiMessage/AiCompleted/DB V39(ai_messages 加 cache_hit/miss/reasoning 列)+ message_repo 映射(持久化)+ 前端 MessageList 显 in·cache·out·reason(in=cache_miss 全价,reasoning 有才显)+ 点击 token 弹详情面板(完整 usage+缓存命中率+model)+ df-miniapp 同步

base前置(提升 prompt cache 命中率):chat.rs aug 拼 base 后(4处)+ knowledge_inject 知识拼 base 后(固定 base 前缀,cache 命中)

附修:replace_conversation 原 13 列 INSERT 丢消息级 token → 改 18 列
This commit is contained in:
lxy
2026-08-03 01:22:30 +08:00
parent 864c696b70
commit a031521776
25 changed files with 563 additions and 45 deletions
Download Patch File Download Diff File Expand all files Collapse all files
crates/df-ai-core/src/provider.rs
+22 -7
View File
@@ -42,25 +42,25 @@ impl ContentPart {
impl ChatMessage {
pub fn system(content: impl Into<String>) -> Self {
Self { id: Some(new_message_id()), role: MessageRole::System, content: content.into(), parts: None, tool_call_id: None, tool_calls: None, model: None, status: None, reasoning_content: None, prompt_tokens: None, completion_tokens: None, timestamp: Some(now_millis_i64()) }
Self { id: Some(new_message_id()), role: MessageRole::System, content: content.into(), parts: None, tool_call_id: None, tool_calls: None, model: None, status: None, reasoning_content: None, prompt_tokens: None, completion_tokens: None, prompt_cache_hit_tokens: None, prompt_cache_miss_tokens: None, reasoning_tokens: None, timestamp: Some(now_millis_i64()) }
}
pub fn user(content: impl Into<String>) -> Self {
Self { id: Some(new_message_id()), role: MessageRole::User, content: content.into(), parts: None, tool_call_id: None, tool_calls: None, model: None, status: None, reasoning_content: None, prompt_tokens: None, completion_tokens: None, timestamp: Some(now_millis_i64()) }
Self { id: Some(new_message_id()), role: MessageRole::User, content: content.into(), parts: None, tool_call_id: None, tool_calls: None, model: None, status: None, reasoning_content: None, prompt_tokens: None, completion_tokens: None, prompt_cache_hit_tokens: None, prompt_cache_miss_tokens: None, reasoning_tokens: None, timestamp: Some(now_millis_i64()) }
}
pub fn assistant(content: impl Into<String>) -> Self {
Self { id: Some(new_message_id()), role: MessageRole::Assistant, content: content.into(), parts: None, tool_call_id: None, tool_calls: None, model: None, status: None, reasoning_content: None, prompt_tokens: None, completion_tokens: None, timestamp: Some(now_millis_i64()) }
Self { id: Some(new_message_id()), role: MessageRole::Assistant, content: content.into(), parts: None, tool_call_id: None, tool_calls: None, model: None, status: None, reasoning_content: None, prompt_tokens: None, completion_tokens: None, prompt_cache_hit_tokens: None, prompt_cache_miss_tokens: None, reasoning_tokens: None, timestamp: Some(now_millis_i64()) }
}
pub fn assistant_with_tools(content: impl Into<String>, tool_calls: Vec<ToolCall>) -> Self {
Self { id: Some(new_message_id()), role: MessageRole::Assistant, content: content.into(), parts: None, tool_call_id: None, tool_calls: Some(tool_calls), model: None, status: None, reasoning_content: None, prompt_tokens: None, completion_tokens: None, timestamp: Some(now_millis_i64()) }
Self { id: Some(new_message_id()), role: MessageRole::Assistant, content: content.into(), parts: None, tool_call_id: None, tool_calls: Some(tool_calls), model: None, status: None, reasoning_content: None, prompt_tokens: None, completion_tokens: None, prompt_cache_hit_tokens: None, prompt_cache_miss_tokens: None, reasoning_tokens: None, timestamp: Some(now_millis_i64()) }
}
pub fn tool_result(call_id: impl Into<String>, content: impl Into<String>) -> Self {
Self { id: Some(new_message_id()), role: MessageRole::Tool, content: content.into(), parts: None, tool_call_id: Some(call_id.into()), tool_calls: None, model: None, status: None, reasoning_content: None, prompt_tokens: None, completion_tokens: None, timestamp: Some(now_millis_i64()) }
Self { id: Some(new_message_id()), role: MessageRole::Tool, content: content.into(), parts: None, tool_call_id: Some(call_id.into()), tool_calls: None, model: None, status: None, reasoning_content: None, prompt_tokens: None, completion_tokens: None, prompt_cache_hit_tokens: None, prompt_cache_miss_tokens: None, reasoning_tokens: None, timestamp: Some(now_millis_i64()) }
}
/// 多模态 user 消息：content 文本 + parts（含 Image 片）。
/// content 作为人类可读文本（也作非 vision 端点降级载荷）；parts 透传给 vision 端点。
pub fn user_parts(content: impl Into<String>, parts: Vec<ContentPart>) -> Self {
Self { id: Some(new_message_id()), role: MessageRole::User, content: content.into(), parts: Some(parts), tool_call_id: None, tool_calls: None, model: None, status: None, reasoning_content: None, prompt_tokens: None, completion_tokens: None, timestamp: Some(now_millis_i64()) }
Self { id: Some(new_message_id()), role: MessageRole::User, content: content.into(), parts: Some(parts), tool_call_id: None, tool_calls: None, model: None, status: None, reasoning_content: None, prompt_tokens: None, completion_tokens: None, prompt_cache_hit_tokens: None, prompt_cache_miss_tokens: None, reasoning_tokens: None, timestamp: Some(now_millis_i64()) }
}
/// 是否含图片片（供 provider 判定走多模态分支）。
@@ -315,6 +315,11 @@ mod tests {
model: None,
status: None,
reasoning_content: None,
prompt_tokens: None,
completion_tokens: None,
prompt_cache_hit_tokens: None,
prompt_cache_miss_tokens: None,
reasoning_tokens: None,
timestamp: None,
};
assert_eq!(m.content, "字面量构造");
@@ -367,6 +372,11 @@ mod tests {
model: None,
status: None,
reasoning_content: Some("thinking process".to_string()),
prompt_tokens: None,
completion_tokens: None,
prompt_cache_hit_tokens: None,
prompt_cache_miss_tokens: None,
reasoning_tokens: None,
timestamp: None,
};
let json = serde_json::to_string(&m).unwrap();
@@ -422,6 +432,11 @@ mod tests {
model: None,
status: None,
reasoning_content: None,
prompt_tokens: None,
completion_tokens: None,
prompt_cache_hit_tokens: None,
prompt_cache_miss_tokens: None,
reasoning_tokens: None,
timestamp: None,
};
let json = serde_json::to_string(&m).unwrap();
@@ -469,7 +484,7 @@ mod tests {
let resp = CompletionResponse {
text: "ok".to_string(),
model: "r1".to_string(),
usage: TokenUsage { prompt_tokens: 10, completion_tokens: 20, total_tokens: 30 },
usage: TokenUsage { prompt_tokens: 10, completion_tokens: 20, total_tokens: 30, ..Default::default() },
tool_calls: None,
reasoning_content: Some("r1 thought".to_string()),
};
crates/df-ai-core/src/types.rs
+35
View File
@@ -124,6 +124,20 @@ pub struct ChatMessage {
/// 语义同 prompt_tokens;provider 流式 usage 缺失时(GLM 等)可能为 0。
#[serde(default, skip_serializing_if = "Option::is_none")]
pub completion_tokens: Option<u32>,
/// 缓存命中 token(低价,deepseek prompt_cache_hit / anthropic cache_read)。
/// token 分项显示(2026-08-02):前端 in=miss(全价真实)+ cache=hit(命中) 分计费展示。
/// 老 JSON 反序列化为 None(向前兼容);非 cache provider 恒为 0。
#[serde(default, skip_serializing_if = "Option::is_none")]
pub prompt_cache_hit_tokens: Option<u32>,
/// 未命中 token(全价真实输入,deepseek prompt_cache_miss / anthropic cache_creation)。
/// 前端 in 显示用此字段(非 prompt_tokens 总,避免掩盖命中比例)。
/// 老 JSON 反序列化为 None(向前兼容)。
#[serde(default, skip_serializing_if = "Option::is_none")]
pub prompt_cache_miss_tokens: Option<u32>,
/// 思考 token(deepseek-reasoner/o1 reasoning_tokens,隐藏输出)。
/// 前端仅 > 0 时显示(reason 后缀);老 JSON 反序列化为 None。
#[serde(default, skip_serializing_if = "Option::is_none")]
pub reasoning_tokens: Option<u32>,
}
/// 当前 Unix 毫秒(ChatMessage 打戳用;df-ai-core 不依赖 df-types,内联避免新增依赖)。
@@ -287,11 +301,32 @@ pub struct CompletionResponse {
}
/// Token 用量
///
/// 分项字段(token 分项显示 + 详情面板,2026-08-02):
/// - `prompt_tokens`:输入总(= cache_hit + cache_miss,兼容老链路保留;前端不再单独展示,
/// 改用 cache_miss 作 in 真实全价消耗)。
/// - `completion_tokens`:输出。
/// - `total_tokens`:总计(各 provider 计费不同,相加无意义,前端不显;保留供老链路/日志)。
/// - `prompt_cache_hit_tokens`:缓存命中(低价,deepseek/anthropic cache_read)。
/// - `prompt_cache_miss_tokens`:未命中(全价真实输入,deepseek/openai 扩展字段)。
/// - `reasoning_tokens`:思考(deepseek-reasoner/o1 隐藏输出 token)。
///
/// 不同 provider 字段名不同,serde default 兜底(无则 0),向前兼容老响应。
/// 构造点全用 `..Default::default()` 补缺,详见各 provider 解析点。
#[derive(Debug, Clone, Default, Serialize, Deserialize)]
pub struct TokenUsage {
pub prompt_tokens: u32,
pub completion_tokens: u32,
pub total_tokens: u32,
/// 缓存命中 token(低价,deepseek prompt_cache_hit / anthropic cache_read_input_tokens)
#[serde(default)]
pub prompt_cache_hit_tokens: u32,
/// 未命中 token(全价真实输入,deepseek prompt_cache_miss / anthropic cache_creation)
#[serde(default)]
pub prompt_cache_miss_tokens: u32,
/// 思考 token(deepseek-reasoner/o1 reasoning_tokens,隐藏输出)
#[serde(default)]
pub reasoning_tokens: u32,
}
/// 流式输出的 chunk