Skip to content

ccproxy.services.adapters.delta_utils

ccproxy.services.adapters.delta_utils

Delta accumulation utilities following OpenAI SDK patterns.

accumulate_delta 

accumulate_delta(accumulated, delta)

Recursively merge delta into accumulated following OpenAI's rules.

This function implements the same accumulation logic as OpenAI's SDK: - Concatenate strings - Add numbers (int/float) - Recursively merge dictionaries - Extend primitive lists - Merge object lists by 'index' key - Preserve 'index' and 'type' keys without modification

Parameters:

Name Type Description Default
accumulated dict[str, Any]

The accumulated state to merge into

 required
delta dict[str, Any]

The delta to merge

 required

Returns:

Type Description
dict[str, Any]

The merged result (may modify accumulated in-place)

Raises:

Type Description
TypeError

For unsupported data types
ValueError

For invalid list structures
Source code in ccproxy/services/adapters/delta_utils.py 
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
def accumulate_delta(
    accumulated: dict[str, Any], delta: dict[str, Any]
) -> dict[str, Any]:
    """Recursively merge delta into accumulated following OpenAI's rules.

    This function implements the same accumulation logic as OpenAI's SDK:
    - Concatenate strings
    - Add numbers (int/float)
    - Recursively merge dictionaries
    - Extend primitive lists
    - Merge object lists by 'index' key
    - Preserve 'index' and 'type' keys without modification

    Args:
        accumulated: The accumulated state to merge into
        delta: The delta to merge

    Returns:
        The merged result (may modify accumulated in-place)

    Raises:
        TypeError: For unsupported data types
        ValueError: For invalid list structures
    """
    # Handle None/empty cases
    if not delta:
        return accumulated
    if not accumulated:
        return dict(delta)

    # Work on a copy to avoid mutating input
    result = dict(accumulated)

    for key, delta_value in delta.items():
        if key not in result:
            # New key, just set it
            result[key] = delta_value
            continue

        if key in ("index", "type", "id", "name", "call_id"):
            # Identity/discriminator fields: overwrite instead of merging.
            #
            # Per OpenAI Chat streaming spec, id/type/function.name only
            # appear on the first tool_call chunk; subsequent chunks carry
            # function.arguments. But the Codex Responses->Chat adapter
            # re-sends these fields on every chunk, so without this branch
            # the generic string-concat branch would produce
            # "shellshell.../fc_abc_fc_abc_..." and break downstream
            # consumers. "index" is included so that non-zero int indices
            # (e.g. index=1) are preserved rather than doubled by the
            # numeric-add branch.
            result[key] = delta_value
            continue

        current_value = result[key]

        # Handle different data type combinations
        if isinstance(current_value, str) and isinstance(delta_value, str):
            # Concatenate strings
            result[key] = current_value + delta_value

        elif isinstance(current_value, int | float) and isinstance(
            delta_value, int | float
        ):
            # Add numbers
            result[key] = current_value + delta_value

        elif isinstance(current_value, dict) and isinstance(delta_value, dict):
            # Recursively merge dictionaries
            result[key] = accumulate_delta(current_value, delta_value)

        elif isinstance(current_value, list) and isinstance(delta_value, list):
            # Handle list merging
            result[key] = _accumulate_list(current_value, delta_value)

        else:
            # For any other case, delta value overwrites
            result[key] = delta_value

    return result