UNPKG

@meldscience/meld

Version:

pipeable one-shot prompt scripting toolkit

118 lines (116 loc) • 4.93 kB
1
2
3
4
5
6
7
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
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
# TREE.md ├── bin/ │ ├── ps.ts │ │ # CLI for `ps` │ │ └── uses PromptScript to process .ps.md files │ │ │ │ # Data Flow: │ │ # 1. Parse CLI arguments for input/output │ │ # 2. Create PromptScript instance │ │ # 3. PromptScript.process() → extracts, executes, replaces │ │ # 4. Writes processed output to file or stdout │ │ │ └── oneshot.ts │ # CLI for `oneshot` │ └── uses Oneshot to send prompts to AI providers │ │ │ # Data Flow: │ # 1. Parse CLI arguments (model, prompt, system prompt, variations, iterations) │ # 2. Select provider (Anthropic or OpenAI) │ # 3. Oneshot.process() → single or multiple calls to chosen AI provider │ # 4. Print or save responses │ │ │ └── oneshotcat.ts │ # CLI for `oneshotcat` │ └── 1-step “ps + oneshot” combo │ │ │ # Data Flow: │ # 1. Parse CLI arguments (model, .ps.md file, etc.) │ # 2. PromptScript.process() to expand commands │ # 3. Oneshot.process() on expanded text │ # 4. Print or save combined response │ ├── src/ │ ├── config.ts │ │ # Loads and merges config from env, .rc file, CLI │ │ # exports loadConfig(options?: ToolConfig): ToolConfig │ │ │ ├── errors.ts │ │ # Centralized error definitions │ │ └── class ToolError │ │ # Extend base Error; includes error codes & details │ │ │ ├── prompt-script.ts │ │ # Core logic for the `ps` tool │ │ └── class PromptScript │ │ # 1) parse .ps.md file │ │ # 2) extract all @cmd[...] occurrences │ │ # 3) execute each command │ │ # 4) replace placeholders with command output in the final text │ │ │ │ │ # Data Flow: │ │ # 1. readFile → extractCommands(content) │ │ # 2. for each command → executeCommand(command) │ │ # 3. replaceCommands(content, results) │ │ # 4. return final content string │ │ │ │ │ ├── constructor(options: PSOptions) │ │ ├── async process(): Promise<string> │ │ ├── async extractCommands(content: string): Promise<Command[]> │ │ ├── async executeCommand(cmd: Command): Promise<CommandResult> │ │ └── async replaceCommands(content: string, results: CommandResult[]): Promise<string> │ │ │ ├── oneshot.ts │ │ # Core logic for the `oneshot` tool │ │ └── class Oneshot │ │ # 1) Reads prompt from file │ │ # 2) Possibly merges with system prompt │ │ # 3) Possibly iterates over variations │ │ # 4) Calls chosen AI provider │ │ # 5) Returns aggregated result │ │ │ │ │ # Data Flow: │ │ # 1. readFile(promptFile) → userPrompt │ │ # 2. for each variation: │ │ # → provider.sendPrompt({ model, systemPrompt, userPrompt }) │ │ # 3. collects responses in an array │ │ # 4. returns results │ │ │ │ │ ├── constructor(options: OneshotOptions, provider: AIProvider) │ │ └── async process(): Promise<ResponseEnvelope[]> │ │ │ ├── oneshotcat.ts │ │ # Combines PromptScript + Oneshot in a single class │ │ └── class Oneshotcat │ │ # 1) Uses PromptScript to expand .ps.md file │ │ # 2) Then uses Oneshot to call AI │ │ │ │ │ # Data Flow: │ │ # 1. promptScript.process() → expandedMarkdown │ │ # 2. oneshot.process() with expandedMarkdown │ │ # 3. return aggregated results │ │ │ │ │ ├── constructor(options: OneshotOptions & PSOptions) │ │ └── async process(): Promise<ResponseEnvelope[]> │ │ │ └── providers/ │ ├── anthropic.ts │ │ # Anthropic-specific AI calls │ │ └── class AnthropicProvider implements AIProvider │ │ # uses anthropicApiKey from config │ │ # calls Anthropic's API endpoint │ │ # returns string │ │ │ └── openai.ts │ # OpenAI-specific AI calls │ └── class OpenAIProvider implements AIProvider │ # uses openaiApiKey from config │ # calls OpenAI's API endpoint │ # returns string │ ├── tests/ │ ├── prompt-script.test.ts │ ├── oneshot.test.ts │ └── oneshotcat.test.ts │ └── package.json