server: streaming of tool calls and thoughts when --jinja is on (#12379)

* add common_json w/ support for truncated json healing

* add common_chat_msg_diff

* partial common_chat_parse

* refactor parser w/ optionals

* server: wire chat diffs in stream mode

* fix trigger of thinking models (must happen after thoughts are closed)

* fix functionary v3.2 raw python!

* rename: common_chat_syntax (now contains format)

* rm common_regex.at_start

* don't return empty <think></think>

* accommodate yet another deepseek r1 distill fantasy syntax (`<｜tool▁calls｜>`)

* fix QwQ 32B tool call parsing after thoughts (hermes2)

* better logs for grammar triggers

* consume spaces after parse_json_tool_calls

* fix required tool calls w/ thinking models that have pre-opened thinking tags

* fix thinking model's initial trigger + test qwq's template

* run most test_tool_call tests in stream + non-stream modes

* make functionary v3.2 parsing more strict (differentiate first match from others)

* send final diff from server, to close off raw python arguments

* support partial content streaming in Generic mode

* tool-call: allow content prelude before hermes2 tool calls (for Qwen2.5)

* Update function-calling.md

* Update tool_bench.py

* chat-parser: remove input from exception (llm output may contain PII)

---------

Co-authored-by: ochafik <ochafik@google.com>
Co-authored-by: Olivier Chafik <ochafik@users.noreply.github.com>

scripts/tool_bench.py

@@ -12,6 +12,7 @@
         export LLAMA_SERVER_BIN_PATH=$PWD/build/bin/llama-server
         export LLAMA_CACHE=${LLAMA_CACHE:-$HOME/Library/Caches/llama.cpp}
 
+        ./scripts/tool_bench.py run --n 10 --temp -1 --temp 0 --temp 1 --temp 2 --temp 5 --llama-baseline $PWD/buildMaster/bin/llama-server --output qwen14b.jsonl --hf bartowski/Qwen2.5-14B-Instruct-GGUF:Q4_K_L
         ./scripts/tool_bench.py run --n 30 --temp -1 --temp 0 --temp 1 --model "Qwen 2.5 1.5B Q4_K_M"      --output qwen1.5b.jsonl  --hf bartowski/Qwen2.5-1.5B-Instruct-GGUF      --ollama qwen2.5:1.5b-instruct-q4_K_M
         ./scripts/tool_bench.py run --n 30 --temp -1 --temp 0 --temp 1 --model "Qwen 2.5 Coder 7B Q4_K_M"  --output qwenc7b.jsonl   --hf bartowski/Qwen2.5-Coder-7B-Instruct-GGUF  --ollama qwen2.5-coder:7b
 
@@ -205,6 +206,7 @@ def run(
     model: Annotated[Optional[str], typer.Option(help="Name of the model to test (server agnostic)")] = None,
     hf: Annotated[Optional[str], typer.Option(help="GGUF huggingface model repo id (+ optional quant) to test w/ llama-server")] = None,
     chat_template: Annotated[Optional[str], typer.Option(help="Chat template override for llama-server")] = None,
+    chat_template_file: Annotated[Optional[str], typer.Option(help="Chat template file override for llama-server")] = None,
     ollama: Annotated[Optional[str], typer.Option(help="Ollama model tag to test")] = None,
     llama_baseline: Annotated[Optional[str], typer.Option(help="llama-server baseline binary path to use as baseline")] = None,
     n: Annotated[int, typer.Option(help="Number of times to run each test")] = 10,
@@ -229,6 +231,12 @@ def run(
     # n_ctx = 8192
     n_ctx = 2048
 
+    if model is None:
+        if hf is not None:
+            model = hf.split("/")[-1]
+        elif ollama is not None:
+            model = ollama
+
     assert force or append or not output.exists(), f"Output file already exists: {output}; use --force to overwrite"
 
     with output.open('a' if append else 'w') as output_file:
@@ -320,6 +328,7 @@ def run(
                     server.model_hf_repo = hf
                     server.model_hf_file = None
                     server.chat_template = chat_template
+                    server.chat_template_file = chat_template_file
                     server.server_path = server_path
                     if port is not None:
                         server.server_port = port
@@ -335,6 +344,7 @@ def run(
                                 temp=t,
                                 output_kwargs=dict(
                                     chat_template=chat_template,
+                                    chat_template_file=chat_template_file,
                                 ),
                                 request_kwargs=dict(
                                     ignore_chat_grammar=ignore_chat_grammar,
@@ -355,6 +365,7 @@ def run(
                         temp=t,
                         output_kwargs=dict(
                             chat_template=None,
+                            chat_template_file=None,
                         ),
                         request_kwargs=dict(
                             model=ollama,