qwen35_adapter.cpp

Go to the documentation of this file.

// SPDX-License-Identifier: Apache-2.0
#include "qwen35_adapter.h"
 
#include "xml_parameter_parser.h"
 
#include <entropic/types/logging.h>
 
#include <nlohmann/json.hpp>
 
#include <regex>
#include <sstream>
 
namespace entropic {
 
namespace {
auto logger = entropic::log::get("inference.adapter.qwen35");
 
constexpr const char* TOOL_RESULT_SUFFIX =
    "Continue. Batch multiple tool calls in one response when possible.";
 
std::string generate_uuid() {
    static std::atomic<uint64_t> counter{0};
    return "tc-" + std::to_string(counter.fetch_add(1, std::memory_order_relaxed));
}
 
} // anonymous namespace
 
// ── Tool call parsing ──────────────────────────────────────
 
ParseResult Qwen35Adapter::parse_tool_calls(const std::string& content) const {
    ParseResult result;
 
    // Primary: XML function calls
    auto calls = parse_xml_function_calls(content);
 
    // Fallback: tagged JSON (base class)
    if (calls.empty()) {
        calls = parse_tagged_tool_calls(content);
    }
 
    result.tool_calls = std::move(calls);
    result.cleaned_content = clean_content(content);
    return result;
}
 
std::vector<ToolCall> Qwen35Adapter::parse_xml_function_calls(
    const std::string& content) const
{
    std::vector<ToolCall> calls;
    std::regex func_pattern(R"(<function=([^>]+)>([\s\S]*?)</function>)");
 
    auto begin = std::sregex_iterator(content.begin(), content.end(), func_pattern);
    auto end = std::sregex_iterator();
 
    for (auto it = begin; it != end; ++it) {
        std::string func_name = (*it)[1].str();
        std::string func_body = (*it)[2].str();
 
        // Trim whitespace from function name
        auto ns = func_name.find_first_not_of(" \t");
        auto ne = func_name.find_last_not_of(" \t");
        if (ns != std::string::npos) {
            func_name = func_name.substr(ns, ne - ns + 1);
        }
 
        auto arguments = extract_xml_parameters(func_body);
 
        ToolCall tc;
        tc.id = generate_uuid();
        tc.name = func_name;
        tc.arguments = std::move(arguments);
        calls.push_back(std::move(tc));
 
        logger->info("Parsed XML tool call: {}", func_name);
    }
    return calls;
}
 
std::unordered_map<std::string, std::string> Qwen35Adapter::extract_xml_parameters(
    const std::string& func_body) const
{
    return entropic::inference::adapters::parse_xml_parameters(
        func_body, logger);
}
 
// ── Tool result formatting ─────────────────────────────────
 
Message Qwen35Adapter::format_tool_result(
    const ToolCall& tool_call,
    const std::string& result) const
{
    Message msg;
    msg.role = "user";
    msg.content = "<tool_response>\n" + result +
                  "\n</tool_response>\n\n" + TOOL_RESULT_SUFFIX;
    return msg;
}
 
// ── Tool definition formatting ─────────────────────────────
 
std::string Qwen35Adapter::format_tools(
    const std::vector<std::string>& tool_jsons) const
{
    nlohmann::json tool_defs = nlohmann::json::array();
 
    for (const auto& json_str : tool_jsons) {
        try {
            auto j = nlohmann::json::parse(json_str);
            tool_defs.push_back({
                {"type", "function"},
                {"function", {
                    {"name", j.value("name", "unknown")},
                    {"description", j.value("description", "")},
                    {"parameters", j.value("inputSchema",
                        nlohmann::json::object())}
                }}
            });
        } catch (...) {
            continue;
        }
    }
 
    std::ostringstream out;
    out << "# Tools\n\n"
        << "You may call one or more functions to assist with the user query.\n"
        << "Put your final answer OUTSIDE of any tool calls.\n\n"
        << "Here are the available tools:\n"
        << "<tools>\n"
        << tool_defs.dump(2) << "\n"
        << "</tools>\n\n"
        << "For each function call, return within <tool_call></tool_call> XML tags:\n"
        << "<tool_call>\n"
        << "<function=example_function>\n"
        << "<parameter=param_name>value</parameter>\n"
        << "</function>\n"
        << "</tool_call>";
    return out.str();
}
 
// ── Content cleaning ───────────────────────────────────────
 
std::string Qwen35Adapter::clean_content(const std::string& content) const {
    // Remove <tool_call>...</tool_call> blocks
    std::string cleaned = std::regex_replace(content,
        std::regex(R"(<tool_call>\s*[\s\S]*?\s*</tool_call>)"), "");
 
    // Remove standalone <function=...>...</function>
    cleaned = std::regex_replace(cleaned,
        std::regex(R"(<function=[^>]+>[\s\S]*?</function>)"), "");
 
    // Strip think blocks
    cleaned = strip_think_blocks(cleaned);
    return cleaned;
}
 
// ── Vision / multimodal (v1.9.11) ──────────────────────────
 
static constexpr const char* VISION_INSTRUCTION =
    "\n\nYou can see and analyze images. When the user shares an "
    "image, describe what you observe before responding to their "
    "question.";
 
std::string Qwen35Adapter::format_system_with_vision(
    const std::string& base_system,
    bool has_vision) const {
    if (!has_vision) {
        return base_system;
    }
    return base_system + VISION_INSTRUCTION;
}
 
std::string Qwen35Adapter::format_content_parts(
    const std::vector<ContentPart>& parts) const {
    // Qwen3.5 uses OpenAI content array format natively —
    // delegate to base class default.
    return ChatAdapter::format_content_parts(parts);
}
 
} // namespace entropic