-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathnode_intermediate.py
More file actions
1118 lines (937 loc) · 49.9 KB
/
node_intermediate.py
File metadata and controls
1118 lines (937 loc) · 49.9 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
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
import os
import yaml, json
from pocketflow import Node, BatchNode
from utils.crawl_github_files import crawl_github_files
from utils.call_llm import call_llm, call_llm_retry
from utils.crawl_local_files import crawl_local_files
from utils.tools import batch_chunks, length_of_tokens, MAX_TOKENS, SPLIT_TOKENS, split_prompt
# Helper to get content for specific file indices
def get_content_for_indices(files_data, indices):
content_map = {}
for i in indices:
if 0 <= i < len(files_data):
path, content = files_data[i]
content_map[f"{i} # {path}"] = (
content # Use index + path as key for context
)
return content_map
class FetchRepo(Node):
def prep(self, shared):
repo_url = shared.get("repo_url")
local_dir = shared.get("local_dir")
project_name = shared.get("project_name")
if not project_name:
# Basic name derivation from URL or directory
if repo_url:
project_name = repo_url.split("/")[-1].replace(".git", "")
else:
project_name = os.path.basename(os.path.abspath(local_dir))
shared["project_name"] = project_name
# Get file patterns directly from shared
include_patterns = shared["include_patterns"]
exclude_patterns = shared["exclude_patterns"]
max_file_size = shared["max_file_size"]
return {
"repo_url": repo_url,
"local_dir": local_dir,
"token": shared.get("github_token"),
"include_patterns": include_patterns,
"exclude_patterns": exclude_patterns,
"max_file_size": max_file_size,
"use_relative_paths": True,
}
def exec(self, prep_res):
if prep_res["repo_url"]:
print(f"Crawling repository: {prep_res['repo_url']}...")
result = crawl_github_files(
repo_url=prep_res["repo_url"],
token=prep_res["token"],
include_patterns=prep_res["include_patterns"],
exclude_patterns=prep_res["exclude_patterns"],
max_file_size=prep_res["max_file_size"],
use_relative_paths=prep_res["use_relative_paths"],
)
else:
print(f"Crawling directory: {prep_res['local_dir']}...")
result = crawl_local_files(
directory=prep_res["local_dir"],
include_patterns=prep_res["include_patterns"],
exclude_patterns=prep_res["exclude_patterns"],
max_file_size=prep_res["max_file_size"],
use_relative_paths=prep_res["use_relative_paths"],
)
# Convert dict to list of tuples: [(path, content), ...]
files_list = list(result.get("files", {}).items())
if len(files_list) == 0:
raise (ValueError("Failed to fetch files"))
print(f"Fetched {len(files_list)} files.")
return files_list
def post(self, shared, prep_res, exec_res):
shared["files"] = (
exec_res # List of (path, content) tuples ('browserbase\\cli.js', "#!/usr/bin/env node\nimport './dist/program.js';")
)
class IdentifyAbstractions(Node):
def prep(self, shared):
files_data = shared["files"]
project_name = shared["project_name"] # Get project name
language = shared.get("language", "english") # Get language
use_cache = shared.get("use_cache", True) # Get use_cache flag, default to True
max_abstraction_num = shared.get(
"max_abstraction_num", 10
) # Get max_abstraction_num, default to 10
# Helper to create context from files, respecting limits (basic example)
def create_llm_context(files_data):
context = ""
context_list = []
file_total_info = []
file_info = [] # Store tuples of (index, path)
tokens_num = 0
for i, (path, content) in enumerate(files_data):
entry = f"--- File Index {i}: {path} ---\n{content}\n\n"
tokens_num += length_of_tokens(entry)
if tokens_num < SPLIT_TOKENS:
context += entry
file_info.append((i, path))
else:
file_info = [(i, path)]
if length_of_tokens(entry) < SPLIT_TOKENS:
file_total_info.append(file_info)
context_list.append(context)
tokens_num = length_of_tokens(entry)
context = entry
else:
content_temp_list = split_prompt(content, max_tokens=SPLIT_TOKENS*0.8)
for item in content_temp_list:
entry_temp = f"--- File Index {i}: {path} ---\n{item}\n\n"
context_list.append(entry_temp)
file_total_info.append(file_info)
if context:
context_list.append(context)
file_total_info.append(file_info)
return context_list, file_total_info # file_info is list of (index, path)
context_list, file_info = create_llm_context(files_data)
return (
context_list,
file_info,
len(files_data),
project_name,
language,
use_cache,
max_abstraction_num,
) # Return all parameters
def exec(self, prep_res):
(
context_list,
file_info,
file_count,
project_name,
language,
use_cache,
max_abstraction_num,
) = prep_res
print(f"Identifying abstractions using LLM...")
# Add language instruction and hints only if not English
language_instruction = ""
name_lang_hint = ""
desc_lang_hint = ""
if language.lower() != "english":
language_instruction = f"IMPORTANT: Generate the `name` and `description` for each abstraction in **{language.capitalize()}** language. Do NOT use English for these fields.\n\n"
name_lang_hint = f" (value in {language.capitalize()})"
desc_lang_hint = f" (value in {language.capitalize()})"
response_result = ""
for i, context in enumerate(context_list):
file_listing_for_prompt = "\n".join(
[f"- {idx} # {path}" for idx, path in file_info[i]]
)
prompt = f"""
For the project `{project_name}`:
Codebase Context:
{context}
Analyze the codebase context.
Identify the top 5–{max_abstraction_num} core abstractions that an intermediate engineer should know first.
For each abstraction, provide:
1. A concise `name`{name_lang_hint}.
2. A precise, intermediate-level `description` (~280–320 words) that focuses on:
- purpose and responsibilities,
- public interface / key inputs & outputs,
- important dependencies or collaborators,
- typical failure modes or edge cases (briefly).
Avoid beginner analogies; prefer concrete, technical language.
3. A list of relevant `file_indices` (integers) using the format `idx # path/comment`.
List of file indices and paths present in the context:
{file_listing_for_prompt}
Format the output as a YAML list of dictionaries:
```yaml
- name: |
Query Processing{name_lang_hint}
description: |
What it is, why it exists, its responsibilities, key inputs/outputs,
notable dependencies, and common pitfalls—written for an intermediate engineer.
file_indices:
- 0 # path/to/file1.py
- 3 # path/to/related.py
- name: |
Query Optimization{name_lang_hint}
description: |
What it is, when to use it, constraints, and trade-offs—no analogies, concise and technical.
file_indices:
- 5 # path/to/another.js
# ... up to {max_abstraction_num} abstractions
```"""
response = call_llm_retry(prompt, {"max_token": MAX_TOKENS-SPLIT_TOKENS-1000})
response_result += response + "\n"
yaml_str = ""
yaml_str_list = response_result.strip().split("```yaml")
for item in yaml_str_list:
for c in item.strip().split("```"):
if c.strip():
yaml_str += c + "\n"
try:
abstractions = yaml.safe_load(yaml_str)
except Exception as e:
if not yaml_str.startswith("Based on the"):
yaml_str = "Based on the codebase context, here are the core abstractions:\n" + yaml_str
abstractions = yaml.safe_load(yaml_str)
else:
print(yaml_str[:30])
raise e
if isinstance(abstractions, dict):
res = []
for key in abstractions.keys():
res.extend(abstractions[key])
abstractions = res
if not isinstance(abstractions, list):
raise ValueError("LLM Output is not a list")
validated_abstractions = []
for item in abstractions:
if not isinstance(item, dict) or not all(
k in item for k in ["name", "description", "file_indices"]
):
print(item)
raise ValueError(f"Missing keys in abstraction item: {item}")
if not isinstance(item["name"], str):
raise ValueError(f"Name is not a string in item: {item}")
if not isinstance(item["description"], str):
raise ValueError(f"Description is not a string in item: {item}")
if not isinstance(item["file_indices"], list):
raise ValueError(f"file_indices is not a list in item: {item}")
# Validate indices
validated_indices = []
for idx_entry in item["file_indices"]:
try:
if isinstance(idx_entry, int):
idx = idx_entry
elif isinstance(idx_entry, str) and "#" in idx_entry:
idx = int(idx_entry.split("#")[0].strip())
elif isinstance(idx_entry, str) and "-" in idx_entry:
idx = int(idx_entry.split("-")[0].strip())
else:
idx = int(str(idx_entry).strip())
if not (0 <= idx < file_count):
raise ValueError(
f"Invalid file index {idx} found in item {item['name']}. Max index is {file_count - 1}."
)
validated_indices.append(idx)
except (ValueError, TypeError):
raise ValueError(
f"Could not parse index from entry: {idx_entry} in item {item['name']} file_count: {file_count-1}"
)
item["files"] = sorted(list(set(validated_indices)))
# Store only the required fields
validated_abstractions.append(
{
"name": item["name"], # Potentially translated name
"description": item[
"description"
], # Potentially translated description
"files": item["files"],
}
)
print(f"Identified {len(validated_abstractions)} abstractions.")
return validated_abstractions
def post(self, shared, prep_res, exec_res):
shared["abstractions"] = (
exec_res # List of {"name": str, "description": str, "files": [int]}
)
class AnalyzeRelationships(Node):
def prep(self, shared):
abstractions = shared[
"abstractions"
] # Now contains 'files' list of indices, name/description potentially translated
files_data = shared["files"]
project_name = shared["project_name"] # Get project name
language = shared.get("language", "english") # Get language
use_cache = shared.get("use_cache", True) # Get use_cache flag, default to True
# Get the actual number of abstractions directly
num_abstractions = len(abstractions)
# Create context with abstraction names, indices, descriptions, and relevant file snippets
context_list = []
context = "Identified Abstractions:\\n"
context_header = "Identified Abstractions:\\n"
token_nums = 0
all_relevant_indices_list = list()
all_relevant_indices = set()
abstraction_info_for_prompt_list = []
abstraction_info_for_prompt = []
for i, abstr in enumerate(abstractions):
# Use 'files' which contains indices directly
file_indices_str = ", ".join(map(str, abstr["files"]))
# Abstraction name and description might be translated already
info_line = f"- Index {i}: {abstr['name']} (Relevant file indices: [{file_indices_str}])\\n Description: {abstr['description']}"
token_nums += length_of_tokens(info_line)
if token_nums < SPLIT_TOKENS * 0.5:
context += info_line + "\\n"
abstraction_info_for_prompt.append(
f"{i} # {abstr['name']}"
) # Use potentially translated name here too
all_relevant_indices.update(abstr["files"])
else:
context_list.append(context)
abstraction_info_for_prompt_list.append(abstraction_info_for_prompt)
all_relevant_indices_list.append(all_relevant_indices)
context = "Identified Abstractions:\\n" + info_line + "\\n"
token_nums = length_of_tokens(context)
abstraction_info_for_prompt = [f"{i} # {abstr['name']}"]
all_relevant_indices = set(abstr["files"])
if abstraction_info_for_prompt:
context_list.append(context)
abstraction_info_for_prompt_list.append(abstraction_info_for_prompt)
all_relevant_indices_list.append(all_relevant_indices)
for context_index, context in enumerate(context_list):
context += "\\nRelevant File Snippets (Referenced by Index and Path):\\n"
relevant_files_content_map = get_content_for_indices(
files_data, sorted(list(all_relevant_indices_list[context_index]))
)
file_context_str = ""
token_nums = 0
for idx_path, content in relevant_files_content_map.items():
entry = f"--- File: {idx_path} ---\\n{content}" + "\\n\\n"
token_nums += length_of_tokens(entry)
if token_nums < SPLIT_TOKENS * 0.3:
file_context_str += entry
else:
context += file_context_str
file_context_str = ""
token_nums = length_of_tokens(context)
if file_context_str:
context += file_context_str
context_list[context_index] = context
return (
context_list,
abstraction_info_for_prompt_list,
num_abstractions, # Pass the actual count
project_name,
language,
use_cache,
) # Return use_cache
def exec(self, prep_res):
(
context_list,
abstraction_info_listing,
num_abstractions, # Receive the actual count
project_name,
language,
use_cache,
) = prep_res # Unpack use_cache
print(f"Analyzing relationships using LLM...")
# Add language instruction and hints only if not English
language_instruction = ""
lang_hint = ""
list_lang_note = ""
if language.lower() != "english":
language_instruction = f"IMPORTANT: Generate the `summary` and relationship `label` fields in **{language.capitalize()}** language. Do NOT use English for these fields.\n\n"
lang_hint = f" (in {language.capitalize()})"
list_lang_note = f" (Names might be in {language.capitalize()})" # Note for the input list
response_result = ""
for c_index, context in enumerate(context_list):
abstraction_listing = "\n".join(abstraction_info_listing[c_index])
prompt = f"""
Based on the following abstractions and relevant code snippets from the project {project_name}:
List of Abstraction Indices and Names{list_lang_note}:
{abstraction_listing}
Context (Abstractions, Descriptions, Code):
{context}
Please provide:
1. A high-level summary of the project for an intermediate engineer. In a few precise sentences, highlight:
- the main system goals and primary data/control flows,
- key extension points or public APIs,
- notable performance, reliability, or operational considerations.
- Use markdown with bold and italic for emphasis, but keep the tone technical rather than introductory.
2. A list (relationships) describing the essential interactions between abstractions. For each relationship, specify:
- from_abstraction: Index of the source abstraction (e.g., 0 # AbstractionName1)
- to_abstraction: Index of the target abstraction (e.g., 1 # AbstractionName2)
- label: A brief, technical label in a few words (e.g., "Calls", "Depends on", "Publishes", "Wraps", "Configures").
Prefer relationships that are evidenced by function calls, parameter passing, event publishing, or composition.
Keep only relationships that matter for understanding the architecture.
IMPORTANT: Make sure EVERY abstraction appears in at least ONE relationship (as source or target).
Format the output as YAML:
summary: |
A concise, technical explanation for an intermediate audience.
Can span multiple lines with **bold** and *italic* for emphasis.
relationships:
- from_abstraction: 0 # AbstractionName1
to_abstraction: 1 # AbstractionName2
label: "Depends on"
- from_abstraction: 2 # AbstractionName3
to_abstraction: 0 # AbstractionName1
label: "Configures"
# ... other relationships
Now, provide the YAML output:
"""
response = call_llm(
prompt, use_cache=(use_cache and self.cur_retry == 0)
) # Use cache only if enabled and not retrying
response_result += response + "\n"
yaml_str = ""
yaml_str_list = response_result.strip().split("```yaml")
for item in yaml_str_list:
for c in item.strip().split("```"):
if c.strip():
yaml_str += c.strip() + "\n"
relationships_data = yaml.safe_load(yaml_str)
if not isinstance(relationships_data, dict) or not all(
k in relationships_data for k in ["summary", "relationships"]
):
raise ValueError(
"LLM output is not a dict or missing keys ('summary', 'relationships')"
)
if not isinstance(relationships_data["summary"], str):
raise ValueError("summary is not a string")
if not isinstance(relationships_data["relationships"], list):
raise ValueError("relationships is not a list")
# Validate relationships structure
validated_relationships = []
for rel in relationships_data["relationships"]:
# Check for 'label' key
if not isinstance(rel, dict) or not all(
k in rel for k in ["from_abstraction", "to_abstraction", "label"]
):
raise ValueError(
f"Missing keys (expected from_abstraction, to_abstraction, label) in relationship item: {rel}"
)
# Validate 'label' is a string
if not isinstance(rel["label"], str):
raise ValueError(f"Relationship label is not a string: {rel}")
# Validate indices
try:
from_idx = int(str(rel["from_abstraction"]).split("#")[0].strip())
to_idx = int(str(rel["to_abstraction"]).split("#")[0].strip())
if not (
0 <= from_idx < num_abstractions and 0 <= to_idx < num_abstractions
):
raise ValueError(
f"Invalid index in relationship: from={from_idx}, to={to_idx}. Max index is {num_abstractions-1}."
)
validated_relationships.append(
{
"from": from_idx,
"to": to_idx,
"label": rel["label"], # Potentially translated label
}
)
except (ValueError, TypeError):
raise ValueError(f"Could not parse indices from relationship: {rel}")
print(
f"Generated project summary and relationship details {len(validated_relationships)}."
)
return {
"summary": relationships_data["summary"], # Potentially translated summary
"details": validated_relationships, # Store validated, index-based relationships with potentially translated labels
}
def post(self, shared, prep_res, exec_res):
# Structure is now {"summary": str, "details": [{"from": int, "to": int, "label": str}]}
# Summary and label might be translated
shared["relationships"] = exec_res
class OrderChapters(Node):
def prep(self, shared):
abstractions = shared["abstractions"] # Name/description might be translated
relationships = shared["relationships"] # Summary/label might be translated
project_name = shared["project_name"] # Get project name
language = shared.get("language", "english") # Get language
use_cache = shared.get("use_cache", True) # Get use_cache flag, default to True
# Prepare context for the LLM
abstraction_info_for_prompt = []
for i, a in enumerate(abstractions):
abstraction_info_for_prompt.append(
f"- {i} # {a['name']}"
) # Use potentially translated name
abstraction_listing = "\n".join(abstraction_info_for_prompt)
# Use potentially translated summary and labels
summary_note = ""
if language.lower() != "english":
summary_note = (
f" (Note: Project Summary might be in {language.capitalize()})"
)
context = f"Project Summary{summary_note}:\n{relationships['summary']}\n\n"
context += "Relationships (Indices refer to abstractions above):\n"
for rel in relationships["details"]:
from_name = abstractions[rel["from"]]["name"]
to_name = abstractions[rel["to"]]["name"]
# Use potentially translated 'label'
context += f"- From {rel['from']} ({from_name}) to {rel['to']} ({to_name}): {rel['label']}\n" # Label might be translated
list_lang_note = ""
if language.lower() != "english":
list_lang_note = f" (Names might be in {language.capitalize()})"
return (
abstraction_listing,
context,
len(abstractions),
project_name,
list_lang_note,
use_cache,
) # Return use_cache
def exec(self, prep_res):
(
abstraction_listing,
context,
num_abstractions,
project_name,
list_lang_note,
use_cache,
) = prep_res # Unpack use_cache
print("Determining chapter order using LLM...")
# No language variation needed here in prompt instructions, just ordering based on structure
# The input names might be translated, hence the note.
prompt = f"""
Given the following project abstractions and their relationships for the project ```` {project_name} ````:
Abstractions (Index # Name){list_lang_note}:
{abstraction_listing}
Context about relationships and project summary:
{context}
If you are going to write an intermediate-level tutorial for ```` {project_name} ````, what is the best order to explain these abstractions, from first to last?
Start with user-facing entry points or public APIs and foundational orchestration layers.
Then cover core services and cross-cutting modules (e.g., state, caching, messaging).
Finally, explain lower-level utilities, adapters, and infrastructure details.
Output requirements (MANDATORY):
- Return **only** a single fenced code block labeled `yaml`. No prose before or after.
- The content **must** be a YAML **list** (not a mapping).
- Include **exactly** one item per abstraction shown above (same count and indices).
- Each item **must** be either:
- an integer index (e.g., `2`), **or**
- a string of the form `idx # AbstractionName` (e.g., `2 # Public API Gateway`).
- Do **not** add any keys like `order:`; do not include comments except inline `# AbstractionName`.
Format example:
```yaml
- 2 # Public API Gateway
- 0 # Core Orchestrator
- 1 # Domain Service (uses Orchestrator)
- 3 # Persistence Layer
- 4 # Adapters/Utilities
Now, provide the YAML output (fenced, and nothing else):
"""
response = call_llm_retry(
prompt, {"max_token": MAX_TOKENS-SPLIT_TOKENS-1000}
) # Use cache only if enabled and not retrying
try:
yaml_str = response.strip().split("```yaml")[1].split("```")[0].strip()
except Exception as e:
with open("tt.txt", "w", encoding="utf-8") as f:
print(response, file=f)
ordered_indices_raw = yaml.safe_load(yaml_str)
if not isinstance(ordered_indices_raw, list):
raise ValueError("LLM output is not a list")
ordered_indices = []
seen_indices = set()
for entry in ordered_indices_raw:
try:
if isinstance(entry, int):
idx = entry
elif isinstance(entry, str) and "#" in entry:
idx = int(entry.split("#")[0].strip())
else:
idx = int(str(entry).strip())
if not (0 <= idx < num_abstractions):
raise ValueError(
f"Invalid index {idx} in ordered list. Max index is {num_abstractions-1}."
)
if idx in seen_indices:
continue
raise ValueError(f"Duplicate index {idx} found in ordered list.")
ordered_indices.append(idx)
seen_indices.add(idx)
except (ValueError, TypeError):
raise ValueError(
f"Could not parse index from ordered list entry: {entry}"
)
# Check if all abstractions are included
if len(ordered_indices) != num_abstractions:
print(f"Ordered list length ({len(ordered_indices)}) does not match number of abstractions ({num_abstractions}). Missing indices: {set(range(num_abstractions)) - seen_indices}")
# raise ValueError(
# f"Ordered list length ({len(ordered_indices)}) does not match number of abstractions ({num_abstractions}). Missing indices: {set(range(num_abstractions)) - seen_indices}"
# )
print(f"Determined chapter order (indices): {ordered_indices}")
return ordered_indices # Return the list of indices
def post(self, shared, prep_res, exec_res):
# exec_res is already the list of ordered indices
shared["chapter_order"] = exec_res # List of indices
class WriteChapters(BatchNode):
def prep(self, shared):
chapter_order = shared["chapter_order"] # List of indices
abstractions = shared[
"abstractions"
] # List of {"name": str, "description": str, "files": [int]}
files_data = shared["files"] # List of (path, content) tuples
project_name = shared["project_name"]
language = shared.get("language", "english")
use_cache = shared.get("use_cache", True) # Get use_cache flag, default to True
self.chapters_written_so_far = (
[]
) # Use instance variable for temporary storage across exec calls
# Create a complete list of all chapters
all_chapters = []
chapter_filenames = {} # Store chapter filename mapping for linking
for i, abstraction_index in enumerate(chapter_order):
if 0 <= abstraction_index < len(abstractions):
chapter_num = i + 1
chapter_name = abstractions[abstraction_index][
"name"
] # Potentially translated name
# Create safe filename (from potentially translated name)
safe_name = "".join(
c if c.isalnum() else "_" for c in chapter_name
).lower()
filename = f"{i+1:02d}_{safe_name}.md"
# Format with link (using potentially translated name)
all_chapters.append(f"{chapter_num}. [{chapter_name}]({filename})")
# Store mapping of chapter index to filename for linking
chapter_filenames[abstraction_index] = {
"num": chapter_num,
"name": chapter_name,
"filename": filename,
}
# Create a formatted string with all chapters
full_chapter_listing = "\n".join(all_chapters)
items_to_process = []
for i, abstraction_index in enumerate(chapter_order):
if 0 <= abstraction_index < len(abstractions):
abstraction_details = abstractions[
abstraction_index
] # Contains potentially translated name/desc
# Use 'files' (list of indices) directly
related_file_indices = abstraction_details.get("files", [])
# Get content using helper, passing indices
related_files_content_map = get_content_for_indices(
files_data, related_file_indices
)
# Get previous chapter info for transitions (uses potentially translated name)
prev_chapter = None
if i > 0:
prev_idx = chapter_order[i - 1]
prev_chapter = chapter_filenames[prev_idx]
# Get next chapter info for transitions (uses potentially translated name)
next_chapter = None
if i < len(chapter_order) - 1:
next_idx = chapter_order[i + 1]
next_chapter = chapter_filenames[next_idx]
items_to_process.append(
{
"chapter_num": i + 1,
"abstraction_index": abstraction_index,
"abstraction_details": abstraction_details, # Has potentially translated name/desc
"related_files_content_map": related_files_content_map,
"project_name": shared["project_name"], # Add project name
"full_chapter_listing": full_chapter_listing, # Add the full chapter listing (uses potentially translated names)
"chapter_filenames": chapter_filenames, # Add chapter filenames mapping (uses potentially translated names)
"prev_chapter": prev_chapter, # Add previous chapter info (uses potentially translated name)
"next_chapter": next_chapter, # Add next chapter info (uses potentially translated name)
"language": language, # Add language for multi-language support
"use_cache": use_cache, # Pass use_cache flag
# previous_chapters_summary will be added dynamically in exec
}
)
else:
print(
f"Warning: Invalid abstraction index {abstraction_index} in chapter_order. Skipping."
)
print(f"Preparing to write {len(items_to_process)} chapters...")
return items_to_process # Iterable for BatchNode
def exec(self, item):
# This runs for each item prepared above
print(item)
abstraction_name = item["abstraction_details"][
"name"
] # Potentially translated name
abstraction_description = item["abstraction_details"][
"description"
] # Potentially translated description
chapter_num = item["chapter_num"]
project_name = item.get("project_name")
language = item.get("language", "english")
use_cache = item.get("use_cache", True) # Read use_cache from item
print(f"Writing chapter {chapter_num} for: {abstraction_name} using LLM...")
# Prepare file context string from the map
# file_context_str = "\n\n".join(
# f"--- File: {idx_path.split('# ')[1] if '# ' in idx_path else idx_path} ---\n{content}"
# for idx_path, content in item["related_files_content_map"].items()
# )
# Get summary of chapters written *before* this one
# Use the temporary instance variable
previous_chapters_summary = "\n---\n".join(self.chapters_written_so_far)
# Add language instruction and context notes only if not English
language_instruction = ""
concept_details_note = ""
structure_note = ""
prev_summary_note = ""
instruction_lang_note = ""
mermaid_lang_note = ""
code_comment_note = ""
link_lang_note = ""
tone_note = ""
if language.lower() != "english":
lang_cap = language.capitalize()
language_instruction = f"IMPORTANT: Write this ENTIRE tutorial chapter in **{lang_cap}**. Some input context (like concept name, description, chapter list, previous summary) might already be in {lang_cap}, but you MUST translate ALL other generated content including explanations, examples, technical terms, and potentially code comments into {lang_cap}. DO NOT use English anywhere except in code syntax, required proper nouns, or when specified. The entire output MUST be in {lang_cap}.\n\n"
concept_details_note = f" (Note: Provided in {lang_cap})"
structure_note = f" (Note: Chapter names might be in {lang_cap})"
prev_summary_note = f" (Note: This summary might be in {lang_cap})"
instruction_lang_note = f" (in {lang_cap})"
mermaid_lang_note = f" (Use {lang_cap} for labels/text if appropriate)"
code_comment_note = f" (Translate to {lang_cap} if possible, otherwise keep minimal English for clarity)"
link_lang_note = (
f" (Use the {lang_cap} chapter title from the structure above)"
)
tone_note = f" (appropriate for {lang_cap} readers)"
prompt_1_token = f"""
{language_instruction}Write an intermediate-level tutorial chapter (in Markdown) for the project {project_name} about the concept: "{abstraction_name}". This is Chapter {chapter_num}.
Concept Details:
Name: {abstraction_name}
Description:
{abstraction_description}
Complete Tutorial Structure:
{item["full_chapter_listing"]}
Context from previous chapters:
{previous_chapters_summary if previous_chapters_summary else "This is the first chapter."}
Instructions for the chapter (Generate content in {language.capitalize()} unless specified otherwise):
Start with a clear heading: # Chapter {chapter_num}: {abstraction_name}.
If this is not the first chapter, begin with a brief transition from the previous chapter{instruction_lang_note}, referencing it with a proper Markdown link using its name{link_lang_note}.
Motivation: Define the main problem this abstraction solves and when to use it. Provide one central use case that anchors the chapter.
Assume the reader knows the language and basic tooling. Focus on:
responsibilities and public interfaces,
key dependencies and collaboration patterns,
typical failure modes and operational concerns,
performance and scalability notes (brief but concrete).
If the abstraction is complex, break it into key concepts. Explain each with concise, technical prose.
Show how to use the abstraction to solve the central use case.
Provide minimal examples with input/output (≤12 lines per block). If longer is necessary, split into small blocks and narrate between them.
Use short code comments to omit unimportant details.
After each code block, add a short explanation of what just happened and why it matters.
“How it works” section:
First give a step-by-step, code-light walkthrough.
Include a compact sequenceDiagram (≤5 participants). If a name has spaces, use participant QP as Query Processing.
“Under the hood”:
Dive into implementation with references to relevant files.
Show short excerpts (again ≤12 lines each), focusing on boundaries, data flow, and decision points.
When referring to other core abstractions, ALWAYS link them using: Chapter Title, based on the complete tutorial structure above.
Use mermaid diagrams (mermaid) when they clarify data flow or lifecycle.
Prefer precise explanations over analogies. Include brief trade-offs and alternatives where helpful.
End with a concise recap of takeaways and a transition to the next chapter. If there is a next chapter, link it as: Next Chapter Title.
Output only the Markdown for this chapter.
"""
file_context_str_list = []
file_context_str_1 = ""
tokens_num = 0
for idx_path, content in item["related_files_content_map"].items():
temp = f"--- File: {idx_path.split('# ')[1] if '# ' in idx_path else idx_path} ---\n{content}"+ "\n\n"
tokens_num += length_of_tokens(temp)
if tokens_num < SPLIT_TOKENS*0.5:
file_context_str_1 += temp
else:
file_context_str_1 = temp
tokens_num = length_of_tokens(temp)
if tokens_num > SPLIT_TOKENS*0.65:
len_temp = len(temp)
count_start = 0
while True:
file_context_str_list.append(temp[count_start:count_start+10000])
count_start += 10000
if len_temp <= count_start:
break
else:
file_context_str_list.append(file_context_str_1)
if file_context_str_1:
tokens_num = length_of_tokens(file_context_str_1)
if tokens_num > SPLIT_TOKENS*0.65:
len_temp = len(file_context_str_1)
count_start = 0
while True:
file_context_str_list.append(file_context_str_1[count_start:count_start+10000])
count_start += 10000
if len_temp <= count_start:
break
else:
file_context_str_list.append(file_context_str_1)
write_content = []
full_chapter_listing = item["full_chapter_listing"]
for file_context_str in file_context_str_list:
if length_of_tokens(prompt_1_token)> MAX_TOKENS * 0.5:
previous_chapters_summary = previous_chapters_summary[-10000:]
print(length_of_tokens(file_context_str), length_of_tokens(prompt_1_token), length_of_tokens(previous_chapters_summary))
prompt = f"""
{language_instruction}Write an intermediate-level tutorial chapter (in Markdown) for the project {project_name} about the concept: "{abstraction_name}". This is Chapter {chapter_num}.
Concept Details:
Name: {abstraction_name}
Description:
{abstraction_description}
Complete Tutorial Structure:
{item["full_chapter_listing"]}
Context from previous chapters:
{previous_chapters_summary if previous_chapters_summary else "This is the first chapter."}
Relevant Code Snippets (Code itself remains unchanged):
{file_context_str if file_context_str else "No specific code snippets provided for this abstraction."}
Instructions for the chapter (Generate content in {language.capitalize()} unless specified otherwise):
Start with a clear heading: # Chapter {chapter_num}: {abstraction_name}.
If this is not the first chapter, begin with a brief transition from the previous chapter{instruction_lang_note}, referencing it with a proper Markdown link using its name{link_lang_note}.
Motivation: Define the main problem this abstraction solves and when to use it. Provide one central use case that anchors the chapter.
Assume the reader knows the language and basic tooling. Focus on:
responsibilities and public interfaces,
key dependencies and collaboration patterns,
typical failure modes and operational concerns,
performance and scalability notes (brief but concrete).
If the abstraction is complex, break it into key concepts. Explain each with concise, technical prose.
Show how to use the abstraction to solve the central use case.
Provide minimal examples with input/output (≤12 lines per block). If longer is necessary, split into small blocks and narrate between them.
Use short code comments to omit unimportant details.
After each code block, add a short explanation of what just happened and why it matters.
“How it works” section:
First give a step-by-step, code-light walkthrough.
Include a compact sequenceDiagram (≤5 participants). If a name has spaces, use participant QP as Query Processing.
“Under the hood”:
Dive into implementation with references to relevant files.
Show short excerpts (again ≤12 lines each), focusing on boundaries, data flow, and decision points.
When referring to other core abstractions, ALWAYS link them using: Chapter Title, based on the complete tutorial structure above.
Use mermaid diagrams (mermaid) when they clarify data flow or lifecycle.
Prefer precise explanations over analogies. Include brief trade-offs and alternatives where helpful.
End with a concise recap of takeaways and a transition to the next chapter. If there is a next chapter, link it as: Next Chapter Title.
Output only the Markdown for this chapter.
"""
chapter_content = call_llm_retry(prompt, {"max_token": MAX_TOKENS-SPLIT_TOKENS-1000}) # Use cache only if enabled and not retrying
actual_heading = f"# Chapter {chapter_num}: {abstraction_name}" # Use potentially translated name
if not chapter_content.strip().startswith(f"# Chapter {chapter_num}"):
# Add heading if missing or incorrect, trying to preserve content
lines = chapter_content.strip().split("\n")
if lines and lines[0].strip().startswith(
"#"
): # If there's some heading, replace it
lines[0] = actual_heading
chapter_content = "\n".join(lines)
else: # Otherwise, prepend it
chapter_content = f"{actual_heading}\n\n{chapter_content}"
write_content.append(chapter_content)
self.chapters_written_so_far.append("\n".join(write_content))
return "\n".join(write_content) # Return the Markdown string (potentially translated)
def post(self, shared, prep_res, exec_res_list):
# exec_res_list contains the generated Markdown for each chapter, in order
shared["chapters"] = exec_res_list
# Clean up the temporary instance variable
del self.chapters_written_so_far
print(f"Finished writing {len(exec_res_list)} chapters.")
class CombineTutorial(Node):
def prep(self, shared):
project_name = shared["project_name"]
output_base_dir = shared.get("output_dir", "output") # Default output dir
output_path = os.path.join(output_base_dir, project_name)
repo_url = shared.get("repo_url") # Get the repository URL
# language = shared.get("language", "english") # No longer needed for fixed strings
# Get potentially translated data
relationships_data = shared[