common : rework gpt-oss parser (#20393)
* common : rework gpt-oss parser * cont : fix gpt-oss tests * cont : add structured output test * cont : rename final to final_msg
This commit is contained in:
Aldehir Rojas
GitHub
parent
fe00a84b4b
commit
5e8910a0db
2 changed files with 56 additions and 113 deletions
109
common/chat.cpp
109
common/chat.cpp
|
|
@ -933,17 +933,12 @@ static common_chat_params common_chat_params_init_gpt_oss(const common_chat_temp
|
||||||
|
|
||||||
// Copy reasoning to the "thinking" field as expected by the gpt-oss template
|
// Copy reasoning to the "thinking" field as expected by the gpt-oss template
|
||||||
auto adjusted_messages = json::array();
|
auto adjusted_messages = json::array();
|
||||||
for (const auto & msg : inputs.messages) {
|
for (auto msg : inputs.messages) {
|
||||||
auto has_reasoning_content = msg.contains("reasoning_content") && msg.at("reasoning_content").is_string();
|
if (msg.contains("reasoning_content") && msg.at("reasoning_content").is_string()) {
|
||||||
auto has_tool_calls = msg.contains("tool_calls") && msg.at("tool_calls").is_array();
|
msg["thinking"] = msg.at("reasoning_content");
|
||||||
|
msg.erase("content");
|
||||||
if (has_reasoning_content && has_tool_calls) {
|
|
||||||
auto adjusted_message = msg;
|
|
||||||
adjusted_message["thinking"] = msg.at("reasoning_content");
|
|
||||||
adjusted_messages.push_back(adjusted_message);
|
|
||||||
} else {
|
|
||||||
adjusted_messages.push_back(msg);
|
|
||||||
}
|
}
|
||||||
|
adjusted_messages.push_back(msg);
|
||||||
}
|
}
|
||||||
|
|
||||||
auto prompt = common_chat_template_direct_apply(tmpl, inputs, /* messages_override= */ adjusted_messages);
|
auto prompt = common_chat_template_direct_apply(tmpl, inputs, /* messages_override= */ adjusted_messages);
|
||||||
|
|
@ -969,45 +964,31 @@ static common_chat_params common_chat_params_init_gpt_oss(const common_chat_temp
|
||||||
"<|channel|>", "<|constrain|>", "<|message|>", "<|start|>", "<|end|>",
|
"<|channel|>", "<|constrain|>", "<|message|>", "<|start|>", "<|end|>",
|
||||||
};
|
};
|
||||||
|
|
||||||
auto has_tools = inputs.tools.is_array() && !inputs.tools.empty();
|
auto has_tools = inputs.tools.is_array() && !inputs.tools.empty();
|
||||||
auto extract_reasoning = inputs.reasoning_format != COMMON_REASONING_FORMAT_NONE;
|
auto has_response_format = !inputs.json_schema.is_null() && inputs.json_schema.is_object();
|
||||||
auto include_grammar = inputs.tool_choice != COMMON_CHAT_TOOL_CHOICE_NONE && has_tools;
|
auto include_grammar = has_response_format || (has_tools && inputs.tool_choice != COMMON_CHAT_TOOL_CHOICE_NONE);
|
||||||
|
|
||||||
auto parser = build_chat_peg_parser([&](common_chat_peg_builder & p) {
|
auto parser = build_chat_peg_parser([&](common_chat_peg_builder & p) {
|
||||||
const std::string END = "<|end|>";
|
auto start = p.rule("start", p.literal("<|start|>assistant"));
|
||||||
const std::string START = "<|start|>";
|
auto end = p.rule("end", p.literal("<|end|>"));
|
||||||
const std::string MESSAGE = "<|message|>";
|
auto content = p.rule("message-content", p.until("<|end|>"));
|
||||||
const std::string CHANNEL = "<|channel|>";
|
auto channel = p.literal("<|channel|>") + (p.literal("commentary") | p.literal("analysis"));
|
||||||
const std::string CONSTRAIN = "<|constrain|>";
|
auto constrain_type = p.chars("[A-Za-z0-9_-]", 1, -1);
|
||||||
const std::string START_ASSISTANT = START + "assistant";
|
|
||||||
const std::string CHANNEL_ANALYSIS = CHANNEL + "analysis";
|
|
||||||
const std::string CHANNEL_COMMENTARY = CHANNEL + "commentary";
|
|
||||||
const std::string CHANNEL_FINAL = CHANNEL + "final";
|
|
||||||
|
|
||||||
auto the_end = END | p.end();
|
auto analysis = p.rule("analysis", p.literal("<|channel|>analysis<|message|>") + p.reasoning(content) + end);
|
||||||
|
auto preamble = p.rule("preamble", p.literal("<|channel|>commentary<|message|>") + p.content(content) + end);
|
||||||
|
auto final_msg = p.rule("final", p.literal("<|channel|>final<|message|>") + p.content(content));
|
||||||
|
auto any = p.rule("any", preamble | analysis);
|
||||||
|
|
||||||
const std::string analysis_header = CHANNEL_ANALYSIS + MESSAGE;
|
if (has_response_format) {
|
||||||
auto segment_content = p.until(END);
|
auto constraint = p.optional(p.space() + p.literal("<|constrain|>") + constrain_type);
|
||||||
auto analysis_segment = extract_reasoning ?
|
auto response_format = p.rule("response-format",
|
||||||
p.literal(analysis_header) + p.reasoning(segment_content) + p.until(END) + the_end :
|
p.literal("<|channel|>final") + constraint + p.literal("<|message|>") +
|
||||||
p.content(analysis_header + p.until(END) + the_end);
|
p.content(p.schema(p.json(), "response-format-schema", inputs.json_schema)));
|
||||||
|
|
||||||
auto channel_header_content = p.until_one_of({ " to=functions.", MESSAGE });
|
return response_format | (analysis + p.zero_or_more(start + analysis) + start + response_format);
|
||||||
auto content_header = p.choice({ p.literal(CHANNEL_COMMENTARY), p.literal(CHANNEL_FINAL) });
|
|
||||||
auto content_segment = p.rule("content-segment", content_header + channel_header_content + MESSAGE +
|
|
||||||
p.content(segment_content) + the_end);
|
|
||||||
|
|
||||||
if (!inputs.json_schema.is_null()) {
|
|
||||||
auto final_header = p.literal(CHANNEL_FINAL);
|
|
||||||
auto constraint = p.optional(p.space() + p.literal(CONSTRAIN) + channel_header_content);
|
|
||||||
return p.optional(analysis_segment) + final_header + constraint + MESSAGE +
|
|
||||||
p.content(p.schema(p.json(), "response-format", inputs.json_schema));
|
|
||||||
}
|
}
|
||||||
|
|
||||||
auto segment = p.optional(START_ASSISTANT + p.space()) + p.choice({ content_segment, analysis_segment });
|
|
||||||
auto contents = p.optional(segment + p.repeat(p.optional(p.space()) + segment, 0, -1)) + p.end();
|
|
||||||
|
|
||||||
// Tool call parser
|
|
||||||
if (has_tools && inputs.tool_choice != COMMON_CHAT_TOOL_CHOICE_NONE) {
|
if (has_tools && inputs.tool_choice != COMMON_CHAT_TOOL_CHOICE_NONE) {
|
||||||
auto tool_choice = p.choice();
|
auto tool_choice = p.choice();
|
||||||
|
|
||||||
|
|
@ -1016,42 +997,37 @@ static common_chat_params common_chat_params_init_gpt_oss(const common_chat_temp
|
||||||
std::string name = function.at("name");
|
std::string name = function.at("name");
|
||||||
const auto & params = function.at("parameters");
|
const auto & params = function.at("parameters");
|
||||||
|
|
||||||
// Tool call can appear as:
|
auto func_name = p.literal(" to=functions.") + p.tool_name(p.literal(name));
|
||||||
// 1. In role header: " to=functions.NAME<|channel|>..."
|
auto constraint = p.optional(p.space() + p.literal("<|constrain|>") + constrain_type);
|
||||||
// 2. In channel: "<|channel|>(analysis|commentary) to=functions.NAME..."
|
|
||||||
auto func_name = p.literal(" to=functions.") + p.tool_name(p.literal(name));
|
|
||||||
|
|
||||||
auto channel = p.literal(CHANNEL_COMMENTARY) | p.literal(CHANNEL_ANALYSIS);
|
|
||||||
auto constraint = p.space() + p.optional(p.literal(CONSTRAIN) + channel_header_content);
|
|
||||||
auto args = p.tool_args(p.schema(p.json(), "tool-" + name + "-schema", params));
|
auto args = p.tool_args(p.schema(p.json(), "tool-" + name + "-schema", params));
|
||||||
|
|
||||||
// Pattern 1: recipient in role header
|
// recipient in role header
|
||||||
// " to=functions.NAME<|channel|>(analysis|commentary)[constraint]<|message|>ARGS"
|
// <|start|>assistant to=functions.NAME<|channel|>(commentary|analysis)[constraint]<|message|>ARGS
|
||||||
auto tool_in_role = p.tool(p.tool_open(func_name + channel) + constraint + MESSAGE + args);
|
auto tool_in_role = p.tool(p.tool_open(func_name + channel + constraint + p.literal("<|message|>")) + args);
|
||||||
|
|
||||||
// Pattern 2: recipient in channel header
|
// recipient in channel header
|
||||||
// "<|channel|>(analysis|commentary) to=functions.NAME[constraint]<|message|>ARGS"
|
// <|channel|>(commentary|analysis) to=functions.NAME[constraint]<|message|>ARGS
|
||||||
auto tool_in_channel = p.tool(channel + p.tool_open(func_name + constraint + MESSAGE) + args);
|
auto tool_in_channel = p.tool(p.tool_open(channel + func_name + constraint + p.literal("<|message|>")) + args);
|
||||||
|
|
||||||
tool_choice |= tool_in_role | tool_in_channel;
|
tool_choice |= p.rule("tool-" + name, tool_in_role | tool_in_channel);
|
||||||
});
|
});
|
||||||
|
|
||||||
auto min_calls = inputs.tool_choice == COMMON_CHAT_TOOL_CHOICE_REQUIRED ? 1 : 0;
|
auto tool_call = p.trigger_rule("tool-call", tool_choice);
|
||||||
auto max_calls = inputs.parallel_tool_calls ? -1 : 1;
|
|
||||||
|
|
||||||
auto role_start = p.optional(p.space() + p.literal(START_ASSISTANT));
|
if (inputs.tool_choice == COMMON_CHAT_TOOL_CHOICE_REQUIRED) {
|
||||||
auto tool_call = p.rule("tool-call", p.repeat(role_start + tool_choice, min_calls, max_calls) + p.end());
|
return tool_call | ( any + p.zero_or_more(start + any) + start + tool_call);
|
||||||
|
}
|
||||||
|
|
||||||
return p.choice({ p.trigger_rule("single-tool", tool_call), p.trigger_rule("tools", p.one_or_more(segment) + tool_call) });
|
return tool_call | final_msg | (any + p.zero_or_more(start + any) + start + (tool_call | final_msg));
|
||||||
}
|
}
|
||||||
|
|
||||||
return contents;
|
return final_msg | (any + p.zero_or_more(start + any) + start + final_msg);
|
||||||
});
|
});
|
||||||
|
|
||||||
data.parser = parser.save();
|
data.parser = parser.save();
|
||||||
|
|
||||||
if (include_grammar) {
|
if (include_grammar) {
|
||||||
data.grammar_lazy = has_tools && inputs.tool_choice == COMMON_CHAT_TOOL_CHOICE_AUTO;
|
data.grammar_lazy = !(has_response_format || (has_tools && inputs.tool_choice == COMMON_CHAT_TOOL_CHOICE_REQUIRED));
|
||||||
data.grammar = build_grammar([&](const common_grammar_builder & builder) {
|
data.grammar = build_grammar([&](const common_grammar_builder & builder) {
|
||||||
foreach_function(inputs.tools, [&](const json & tool) {
|
foreach_function(inputs.tools, [&](const json & tool) {
|
||||||
const auto & function = tool.at("function");
|
const auto & function = tool.at("function");
|
||||||
|
|
@ -1062,10 +1038,9 @@ static common_chat_params common_chat_params_init_gpt_oss(const common_chat_temp
|
||||||
});
|
});
|
||||||
|
|
||||||
data.grammar_triggers = {
|
data.grammar_triggers = {
|
||||||
{ COMMON_GRAMMAR_TRIGGER_TYPE_PATTERN, "^(?:<\\|start\\|>assistant\\s*)?(\\s+to=functions)" },
|
{ COMMON_GRAMMAR_TRIGGER_TYPE_PATTERN, "^\\s+to$" },
|
||||||
{ COMMON_GRAMMAR_TRIGGER_TYPE_PATTERN, "(?:<\\|end\\|>)(?:<\\|start\\|>assistant\\s*)?(\\s+to=functions)" },
|
{ COMMON_GRAMMAR_TRIGGER_TYPE_PATTERN, "<\\|start\\|>assistant(\\s+to)" },
|
||||||
{ COMMON_GRAMMAR_TRIGGER_TYPE_PATTERN,
|
{ COMMON_GRAMMAR_TRIGGER_TYPE_PATTERN, "<\\|start\\|>assistant(<\\|channel\\|>(?:commentary|analysis)\\s+to)" }
|
||||||
"(?:<\\|start\\|>assistant\\s*)?(<\\|channel\\|>(?:commentary|analysis)\\s+to=functions)" }
|
|
||||||
};
|
};
|
||||||
}
|
}
|
||||||
|
|
||||||
|
|
|
||||||
|
|
@ -2448,7 +2448,7 @@ static void test_template_output_peg_parsers(bool detailed_debug) {
|
||||||
|
|
||||||
// Analysis channel (reasoning) with final channel (content)
|
// Analysis channel (reasoning) with final channel (content)
|
||||||
tst.test(
|
tst.test(
|
||||||
"<|channel|>analysis<|message|>I'm\nthinking<|end|>\n<|channel|>final<|message|>Hello, world!\nWhat's "
|
"<|channel|>analysis<|message|>I'm\nthinking<|end|><|start|>assistant<|channel|>final<|message|>Hello, world!\nWhat's "
|
||||||
"up?")
|
"up?")
|
||||||
.reasoning_format(COMMON_REASONING_FORMAT_AUTO)
|
.reasoning_format(COMMON_REASONING_FORMAT_AUTO)
|
||||||
.expect(message_assist_thoughts)
|
.expect(message_assist_thoughts)
|
||||||
|
|
@ -2461,15 +2461,6 @@ static void test_template_output_peg_parsers(bool detailed_debug) {
|
||||||
.expect_reasoning("I'm\nthinking")
|
.expect_reasoning("I'm\nthinking")
|
||||||
.run();
|
.run();
|
||||||
|
|
||||||
// Reasoning format none - reasoning stays in content
|
|
||||||
tst.test(
|
|
||||||
"<|channel|>analysis<|message|>I'm\nthinking<|end|>\n<|channel|>final<|message|>Hello, world!\nWhat's "
|
|
||||||
"up?")
|
|
||||||
.reasoning_format(COMMON_REASONING_FORMAT_NONE)
|
|
||||||
.expect_content(
|
|
||||||
"<|channel|>analysis<|message|>I'm\nthinking<|end|>Hello, world!\nWhat's up?")
|
|
||||||
.run();
|
|
||||||
|
|
||||||
// Tool call with recipient in role header: " to=functions.NAME<|channel|>analysis<|message|>JSON"
|
// Tool call with recipient in role header: " to=functions.NAME<|channel|>analysis<|message|>JSON"
|
||||||
tst.test(" to=functions.special_function<|channel|>analysis<|message|>{\"arg1\": 1}")
|
tst.test(" to=functions.special_function<|channel|>analysis<|message|>{\"arg1\": 1}")
|
||||||
.tools({ special_function_tool })
|
.tools({ special_function_tool })
|
||||||
|
|
@ -2496,37 +2487,16 @@ static void test_template_output_peg_parsers(bool detailed_debug) {
|
||||||
|
|
||||||
// Tool call with reasoning + content (analysis first, then tool call)
|
// Tool call with reasoning + content (analysis first, then tool call)
|
||||||
tst.test(
|
tst.test(
|
||||||
"<|channel|>analysis<|message|>I'm\nthinking<|end|>\n"
|
"<|channel|>analysis<|message|>I'm\nthinking<|end|>"
|
||||||
"<|start|>assistant to=functions.special_function<|channel|>analysis<|message|>{\"arg1\": 1}")
|
"<|start|>assistant to=functions.special_function<|channel|>analysis<|message|>{\"arg1\": 1}")
|
||||||
.reasoning_format(COMMON_REASONING_FORMAT_AUTO)
|
.reasoning_format(COMMON_REASONING_FORMAT_AUTO)
|
||||||
.tools({ special_function_tool })
|
.tools({ special_function_tool })
|
||||||
.expect(message_assist_call_thoughts)
|
.expect(message_assist_call_thoughts)
|
||||||
.run();
|
.run();
|
||||||
|
|
||||||
// Tool calling with extra channel before
|
// Complex tool calling
|
||||||
tst.test(
|
tst.test(
|
||||||
"<|channel|>analysis<|message|>I'm\nthinking<|end|><|start|>assistant<|channel|>commentary"
|
"<|channel|>analysis<|message|>Thinking about edit...<|end|>"
|
||||||
" to=functions.special_function <|message|>{\"arg1\": 1}")
|
|
||||||
.reasoning_format(COMMON_REASONING_FORMAT_AUTO)
|
|
||||||
.tools({ special_function_tool })
|
|
||||||
.expect(message_assist_call_thoughts)
|
|
||||||
.run();
|
|
||||||
|
|
||||||
// Reasoning after final channel
|
|
||||||
// Tool calling after final channel
|
|
||||||
tst.test(
|
|
||||||
"<|channel|>final<|message|><|end|>"
|
|
||||||
"<|start|>assistant<|channel|>analysis<|message|>Thinking about edit..."
|
|
||||||
)
|
|
||||||
.reasoning_format(COMMON_REASONING_FORMAT_AUTO)
|
|
||||||
.expect_reasoning("Thinking about edit...")
|
|
||||||
.expect_content("")
|
|
||||||
.run();
|
|
||||||
|
|
||||||
// Tool calling after final channel
|
|
||||||
tst.test(
|
|
||||||
"<|channel|>final<|message|><|end|>"
|
|
||||||
"<|start|>assistant<|channel|>analysis<|message|>Thinking about edit...<|end|>"
|
|
||||||
"<|start|>assistant<|channel|>commentary to=functions.edit <|constrain|>json"
|
"<|start|>assistant<|channel|>commentary to=functions.edit <|constrain|>json"
|
||||||
"<|message|>{\"oldString\": \"if (part < railCount - 1) {\", \"newString\": \"if (part < 4) {\", \"replaceAll\": false}"
|
"<|message|>{\"oldString\": \"if (part < railCount - 1) {\", \"newString\": \"if (part < 4) {\", \"replaceAll\": false}"
|
||||||
)
|
)
|
||||||
|
|
@ -2561,19 +2531,17 @@ static void test_template_output_peg_parsers(bool detailed_debug) {
|
||||||
})
|
})
|
||||||
.run();
|
.run();
|
||||||
|
|
||||||
// Parallel tool calls
|
// Structured output
|
||||||
tst.test(
|
tst.test(
|
||||||
" to=functions.special_function<|channel|>analysis<|message|>{\"arg1\": 1}\n"
|
"<|channel|>analysis<|message|>I need to output the invoice details in JSON<|end|>"
|
||||||
"<|start|>assistant to=functions.special_function_with_opt<|channel|>analysis<|message|>{\"arg1\": 1, "
|
"<|start|>assistant<|channel|>final <|constrain|>json"
|
||||||
"\"arg2\": 2}")
|
"<|message|>"
|
||||||
.parallel_tool_calls(true)
|
R"({"amount": 123.45, "date": "2025-12-03"})"
|
||||||
.tools({
|
)
|
||||||
special_function_tool, special_function_tool_with_optional_param
|
.reasoning_format(COMMON_REASONING_FORMAT_AUTO)
|
||||||
})
|
.json_schema(invoice_schema)
|
||||||
.expect_tool_calls({
|
.expect_reasoning("I need to output the invoice details in JSON")
|
||||||
{ "special_function", R"({"arg1": 1})", {} },
|
.expect_content(R"({"amount": 123.45, "date": "2025-12-03"})")
|
||||||
{ "special_function_with_opt", R"({"arg1": 1, "arg2": 2})", {} },
|
|
||||||
})
|
|
||||||
.run();
|
.run();
|
||||||
}
|
}
|
||||||
|
|
||||||
|
|
|
||||||
Loading…
Add table
Add a link
Reference in a new issue