abstractcore 2.3.5__py3-none-any.whl → 2.3.7__py3-none-any.whl

{abstractcore-2.3.5.dist-info → abstractcore-2.3.7.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: abstractcore
-Version: 2.3.5
+Version: 2.3.7
 Summary: Unified interface to all LLM providers with essential infrastructure for tool calling, streaming, and model management
 Author: Laurent-Philippe Albou
 Maintainer: Laurent-Philippe Albou
@@ -254,6 +254,117 @@ python -m abstractllm.utils.cli --provider anthropic --model claude-3-5-haiku-la
 - Quick prototyping with different models
 - Learning AbstractCore capabilities
 
+## Built-in Applications (Ready-to-Use CLI Tools)
+
+AbstractCore includes **three specialized command-line applications** for common LLM tasks. These are production-ready tools that can be used directly from the terminal without any Python programming.
+
+### Available Applications
+
+| Application | Purpose | Direct Command |
+|-------------|---------|----------------|
+| **Summarizer** | Document summarization | `summarizer` |
+| **Extractor** | Entity and relationship extraction | `extractor` |
+| **Judge** | Text evaluation and scoring | `judge` |
+
+### Quick Usage Examples
+
+```bash
+# Document summarization with different styles and lengths
+summarizer document.pdf --style executive --length brief
+summarizer report.txt --focus "technical details" --output summary.txt
+summarizer large_doc.txt --chunk-size 15000 --provider openai --model gpt-4o-mini
+
+# Entity extraction with various formats and options
+extractor research_paper.pdf --format json-ld --focus technology
+extractor article.txt --entity-types person,organization,location --output entities.jsonld
+extractor doc.txt --iterate 3 --mode thorough --verbose
+
+# Text evaluation with custom criteria and contexts
+judge essay.txt --criteria clarity,accuracy,coherence --context "academic writing"
+judge code.py --context "code review" --format plain --verbose
+judge proposal.md --custom-criteria has_examples,covers_risks --output assessment.json
+```
+
+### Installation & Setup
+
+Apps are automatically available after installing AbstractCore:
+
+```bash
+# Install with all features
+pip install abstractcore[all]
+
+# Apps are immediately available
+summarizer --help
+extractor --help  
+judge --help
+```
+
+### Alternative Usage Methods
+
+```bash
+# Method 1: Direct commands (recommended)
+summarizer document.txt
+extractor report.pdf
+judge essay.md
+
+# Method 2: Via Python module
+python -m abstractllm.apps summarizer document.txt
+python -m abstractllm.apps extractor report.pdf
+python -m abstractllm.apps judge essay.md
+```
+
+### Key Parameters
+
+**Common Parameters (all apps):**
+- `--provider` + `--model` - Use different LLM providers (OpenAI, Anthropic, Ollama, etc.)
+- `--output` - Save results to file instead of console
+- `--verbose` - Show detailed progress information
+- `--timeout` - HTTP timeout for LLM requests (default: 300s)
+
+**Summarizer Parameters:**
+- `--style` - Summary style: `structured`, `narrative`, `objective`, `analytical`, `executive`, `conversational`
+- `--length` - Summary length: `brief`, `standard`, `detailed`, `comprehensive`
+- `--focus` - Specific focus area for summarization
+- `--chunk-size` - Chunk size for large documents (1000-32000, default: 8000)
+
+**Extractor Parameters:**
+- `--format` - Output format: `json-ld`, `triples`, `json`, `yaml`
+- `--entity-types` - Focus on specific entities: `person,organization,location,technology,etc.`
+- `--mode` - Extraction mode: `fast`, `balanced`, `thorough`
+- `--iterate` - Number of refinement iterations (1-10, default: 1)
+- `--minified` - Output compact JSON without indentation
+
+**Judge Parameters:**
+- `--context` - Evaluation context (e.g., "code review", "academic writing")
+- `--criteria` - Standard criteria: `clarity,soundness,effectiveness,etc.`
+- `--custom-criteria` - Custom evaluation criteria
+- `--format` - Output format: `json`, `plain`, `yaml`
+- `--include-criteria` - Include detailed criteria explanations
+
+### Key Features
+
+- **Provider Agnostic**: Works with any configured LLM provider (OpenAI, Anthropic, Ollama, etc.)
+- **Multiple Formats**: Support for PDF, TXT, MD, DOCX, and more
+- **Flexible Output**: JSON, JSON-LD, YAML, plain text formats
+- **Batch Processing**: Process multiple files at once
+- **Configurable**: Custom prompts, criteria, and evaluation rubrics
+- **Production Ready**: Robust error handling and logging
+
+### Full Documentation
+
+Each application has comprehensive documentation with examples and advanced usage:
+
+- **[Summarizer Guide](docs/apps/basic-summarizer.md)** - Document summarization with multiple strategies
+- **[Extractor Guide](docs/apps/basic-extractor.md)** - Entity and relationship extraction
+- **[Judge Guide](docs/apps/basic-judge.md)** - Text evaluation and scoring systems
+
+**When to use the apps:**
+- Processing documents without writing code
+- Batch text analysis workflows
+- Quick prototyping of text processing pipelines
+- Integration with shell scripts and automation
+- Standardized text processing tasks
+
 ## Documentation
 
 **📚 Complete Documentation:** [docs/](docs/) - Full documentation index and navigation guide

{abstractcore-2.3.5.dist-info → abstractcore-2.3.7.dist-info}/RECORD

@@ -1,10 +1,10 @@
-abstractcore-2.3.5.dist-info/licenses/LICENSE,sha256=PI2v_4HMvd6050uDD_4AY_8PzBnu2asa3RKbdDjowTA,1078
-abstractllm/__init__.py,sha256=qoh8JJY02pFEGDkTHGiQlWjG5VBMmYn5j8pNBFxTxdE,1839
+abstractcore-2.3.7.dist-info/licenses/LICENSE,sha256=PI2v_4HMvd6050uDD_4AY_8PzBnu2asa3RKbdDjowTA,1078
+abstractllm/__init__.py,sha256=UWHXV-OyMpDeBsCFfLTTgVqkFTrj-uQWVqgFtkfjtxU,1859
 abstractllm/apps/__init__.py,sha256=H6fOR28gyBW8bDCEAft2RUZhNmRYA_7fF91szKBjhpA,30
-abstractllm/apps/__main__.py,sha256=qul2xxC59G_ors-OeuODwfq6XY12ahyEe9GGLmrlHjk,1241
-abstractllm/apps/extractor.py,sha256=MsRgEShvsDYt-k30LdIxzSdcQh_TJGBLK7MFKOKb_4c,22682
-abstractllm/apps/judge.py,sha256=QgvzhwmhUgoFgLPuxB0l90DYtDpyLamSdp4V4zDSKOU,22091
-abstractllm/apps/summarizer.py,sha256=NYU2xjHbwtry0vdPUNbdZDW2xOaSxQAMLXxFxKlp8eo,12477
+abstractllm/apps/__main__.py,sha256=YyAHo-0O4yAJ5JeinsioIM6Ytam2i4eq6UQt2oWB2s0,1579
+abstractllm/apps/extractor.py,sha256=T_9-GTMkTbswa2m3E294IO1eGiedNAWi0dK1xgdm0WE,23491
+abstractllm/apps/judge.py,sha256=gQBpwXLVLIyQxTwHblIPbNWRk6IGsRqkWq2dgj2vO1s,23278
+abstractllm/apps/summarizer.py,sha256=zf3MCJu5mwkOLhTxo41lHuc-462YkUYcFDB5pQu8Uv4,13786
 abstractllm/architectures/__init__.py,sha256=-4JucAM7JkMWShWKkePoclxrUHRKgaG36UTguJihE0U,1046
 abstractllm/architectures/detection.py,sha256=PYYzJ73tXI5ssJOpEL6-mHBmTreYXIq5f4nsfqtrpmw,9465
 abstractllm/architectures/enums.py,sha256=9vIv2vDBEKhxwzwH9iaSAyf-iVj3p8y9loMeN_mYTJ8,3821
@@ -14,47 +14,49 @@ abstractllm/assets/session_schema.json,sha256=b6HTAWxRVlVhAzA7FqaKpunK1yO6jilBOs
 abstractllm/core/__init__.py,sha256=cdAYflvOMUFfiQDV7l972TN95zVXVkr5SixsZ6bNNEw,682
 abstractllm/core/enums.py,sha256=zXDTnMvPPhN3DSqUqg09f-HxTqDbVYn9FGOT-n5LwzA,694
 abstractllm/core/factory.py,sha256=1MSBTqm_sq0PaRqQC4L-U2OzPBkqaKEfg1rjO2uXukc,5234
-abstractllm/core/interface.py,sha256=UadCfs79DuMgkC_7nk5JuoAnmfUvneRqs1Jnr7_r9TE,14367
+abstractllm/core/interface.py,sha256=1YuCu7Q1MYVvQ_1wM1A5cQvhoB82fZku0p5ZapX9DYM,13690
 abstractllm/core/retry.py,sha256=2vTeUQa2_IaMksyHjXWdLvVXXrgPZjXsv2LvANKHd94,14542
-abstractllm/core/session.py,sha256=ZB1XsEFd1wrGAUbBmR2VrLRxR0FxOk1e1uc1bCfIyJE,33864
+abstractllm/core/session.py,sha256=B7gbzLhISOlrO0A_V9m1W3PojqNE6kupii8Es7T-1HE,35194
 abstractllm/core/types.py,sha256=d8Nph272ZdSxg2RqeUgAdefaE5ruRNdGOs0DgM4i1rc,3641
 abstractllm/embeddings/__init__.py,sha256=EES0BQht_zBgxCp5fnBYVLWN2i-YxZQcD2Dyit1varo,641
-abstractllm/embeddings/manager.py,sha256=TsMHFsdIjMI9nEY17wz-5rMkruuDb2ZEXsm5mwEEbAg,50100
+abstractllm/embeddings/manager.py,sha256=YCW1r7ldd_mZshRjWNiNLIyhZktFl1SIiwgDDf41ve4,50043
 abstractllm/embeddings/models.py,sha256=bsPAzL6gv57AVii8O15PT0kxfwRkOml3f3njJN4UDi4,4874
 abstractllm/events/__init__.py,sha256=Z7WIDjerS3G6zjcjiRH3tHIbKdFNzlWnibHMNxkWWzo,11064
 abstractllm/processing/__init__.py,sha256=t6hiakQjcZROT4pw9ZFt2q6fF3vf5VpdMKG2EWlsFd8,540
-abstractllm/processing/basic_extractor.py,sha256=DExf-Sz0Lf2IMxgo8QdlYoD9GyECT6QaMPEgzCpMTDY,36794
-abstractllm/processing/basic_judge.py,sha256=5SHpHD8GhgR7BPO8dm8nfnp8bDwGwN_DmfHLrfRUMWA,29724
-abstractllm/processing/basic_summarizer.py,sha256=rl1IOwOxdgUkHXGN_S_hcXEUYsHQogRj1-wi94jM_28,22631
+abstractllm/processing/basic_extractor.py,sha256=geB9_bqH9Z3hcKq7SheHtsqv0sgitolyMo-SPGgR2Cc,49673
+abstractllm/processing/basic_judge.py,sha256=pHfTSYRjNZq4N3BdpST-hpHgUu6PJFgkeuRo9bK-fiw,31879
+abstractllm/processing/basic_summarizer.py,sha256=P-tkTm0qMDLudQrctAAf1MYqZIa-hvDRZRg8isaVg1U,25334
 abstractllm/providers/__init__.py,sha256=UTpR2Bf_ICFG7M--1kxUmNXs4gl026Tp-KI9zJlvMKU,574
 abstractllm/providers/anthropic_provider.py,sha256=BM8Vu89c974yicvFwlsJ5C3N0wR9Kkt1pOszViWCwAQ,19694
 abstractllm/providers/base.py,sha256=rDxd5XglfZuZNj94iwjWc3ItPSddMQ92Y2G624mb60M,42780
 abstractllm/providers/huggingface_provider.py,sha256=4u3-Z4txonlKXE3DazvgUbxtzVfmdk-ZHsKPSaIfwD4,40785
 abstractllm/providers/lmstudio_provider.py,sha256=HcMltGyiXrLb2acA_XE4wDi1d-X2VZiT3eV1IjMF150,15701
-abstractllm/providers/mlx_provider.py,sha256=_tYg7AhPSsUfCUMqh7FJ--F3ZWr2BNNCtnvH4QtxttU,13745
-abstractllm/providers/mock_provider.py,sha256=5N8Q1alSYezZX-IpurfOA6o5XzKk5xdUjWetz5ntEX4,4060
-abstractllm/providers/ollama_provider.py,sha256=UUatKMguSrhQeuUm2e7Yw64qqaGraSRYPETHPqBcLU8,19277
+abstractllm/providers/mlx_provider.py,sha256=qusbTNvwWB8ydKVwv0adCLpMLFVT2RvX4LYHW-jUxlQ,14166
+abstractllm/providers/mock_provider.py,sha256=qORJxy_Yk0X970GiQG4JfU-T2IUx6C_PNASKM-M7ya8,4597
+abstractllm/providers/ollama_provider.py,sha256=SkXD5gjuzeu9Thqnt4pRPSi-cjWxwuZGV2x5YMm26jo,19340
 abstractllm/providers/openai_provider.py,sha256=xGFrSkbCrsBnWnho1U2aMCBdzfCqf121wU1EFGmU3YQ,21678
 abstractllm/providers/streaming.py,sha256=KQUH1DB-7UvuZP0_Wrcajj1P-9jdqnQfpIEeSYkxVFU,31153
 abstractllm/server/__init__.py,sha256=1DSAz_YhQtnKv7sNi5TMQV8GFujctDOabgvAdilQE0o,249
-abstractllm/server/app.py,sha256=uiaSlqLn9FFVycsVSPPkwDs_EeqBnsLxxhQaWHiu3SY,44243
+abstractllm/server/app.py,sha256=lz6ys1rd2AmsQarwkcOxpLsYQ5xKsp0iskTiYKIC92s,44495
 abstractllm/structured/__init__.py,sha256=hvlV-59u7h7wjavi50C4Fn_Z4UbEsvuF_NG-DruWUxA,338
-abstractllm/structured/handler.py,sha256=H4rxrcpUMBgDbaLpWcVveHkjKp-55JkBL8KcBUXSPNY,15029
+abstractllm/structured/handler.py,sha256=Vb15smiR81JGDXX2RLkY2Exuj67J7a6C-xwVrZoXp0I,17134
 abstractllm/structured/retry.py,sha256=BN_PvrWybyU1clMy2cult1-TVxFSMaVqiCPmmXvA5aI,3805
 abstractllm/tools/__init__.py,sha256=ZTC8V5tSR5ZXbaQdDh4JzTf-jokknazWOYFjrp1NpGw,2257
-abstractllm/tools/common_tools.py,sha256=EjmyJiMglDxNO98gGV9iStts7rn3ZovNqgVzTleiqME,64109
+abstractllm/tools/common_tools.py,sha256=DfscdezYFdRiCVrFBwrcJxEaNJWA8aT-KEcuDq6W10I,64111
 abstractllm/tools/core.py,sha256=lUUGihyceiRYlKUFvEMig9jWFF563d574mSDbYYD3fM,4777
 abstractllm/tools/handler.py,sha256=GmDenXAJkhceWSGlhvuF90aMb2301tRTh6WxGwBQifc,12022
 abstractllm/tools/parser.py,sha256=1r5nmEEp1Rid3JU6ct-s3lP-eCln67fvXG5HCjqiRew,27740
 abstractllm/tools/registry.py,sha256=cN3nbPEK6L2vAd9740MIFf2fitW_4WHpQfK4KvQjnT0,9059
 abstractllm/tools/syntax_rewriter.py,sha256=c3NSTvUF3S3ho5Cwjp7GJJdibeYAI6k3iaBwhKcpTfo,17318
 abstractllm/tools/tag_rewriter.py,sha256=UGFMBj2QKwm12j8JQ6oc2C_N3ZeNqz9Enz4VkEIrS0c,20338
-abstractllm/utils/__init__.py,sha256=245wfpreb4UHXsCQBHKe9DbtjKNo-KIi-k14apBRj_Q,202
-abstractllm/utils/cli.py,sha256=5Dgyk6d17SKHX52n0VWa9dkfd5qsGh4jO44DZjYhsDM,55782
+abstractllm/utils/__init__.py,sha256=M1wyIf6XuqjdPySkODU5M35b8UiIiFbfnZiq5pKAqEc,538
+abstractllm/utils/cli.py,sha256=PcA5AKvAr3bADT9x9OytfE4MDUVh48q8DcbIu9CLkj8,56875
 abstractllm/utils/self_fixes.py,sha256=QEDwNTW80iQM4ftfEY3Ghz69F018oKwLM9yeRCYZOvw,5886
 abstractllm/utils/structured_logging.py,sha256=2_bbMRjOvf0gHsRejncel_-PrhYUsOUySX_eaPcQopc,15827
-abstractcore-2.3.5.dist-info/METADATA,sha256=3BmiLAm1e6EmpJ7QNvmqs0QFzVnAsbDx7PGqrG-Phkg,14987
-abstractcore-2.3.5.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-abstractcore-2.3.5.dist-info/entry_points.txt,sha256=plz04HNXbCbQkmEQj_8xKWW8x7zBxus68HD_04-IARM,306
-abstractcore-2.3.5.dist-info/top_level.txt,sha256=Md-8odCjB7hTNnE5hucnifAoLrL9HvRPffZmCq2jpoI,12
-abstractcore-2.3.5.dist-info/RECORD,,
+abstractllm/utils/token_utils.py,sha256=wAZpqiPh4eL8ppl1lNjqTzxCmlvdJ1Xu62UJyaDq09Y,21096
+abstractllm/utils/version.py,sha256=QyoMwQlCZK5rKvZLjbmcREXZU6Z8CC4a29TpUO4do04,1386
+abstractcore-2.3.7.dist-info/METADATA,sha256=lcnxke_jx-tKffNdv6UM5r45KRHN-YWt9Uos3Zu1gpc,19399
+abstractcore-2.3.7.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+abstractcore-2.3.7.dist-info/entry_points.txt,sha256=plz04HNXbCbQkmEQj_8xKWW8x7zBxus68HD_04-IARM,306
+abstractcore-2.3.7.dist-info/top_level.txt,sha256=Md-8odCjB7hTNnE5hucnifAoLrL9HvRPffZmCq2jpoI,12
+abstractcore-2.3.7.dist-info/RECORD,,

abstractllm/__init__.py

@@ -1,3 +1,4 @@
+# -*- coding: utf-8 -*-
 """
 AbstractLLM - Unified interface to all LLM providers with essential infrastructure.
 
@@ -24,7 +25,7 @@ Quick Start:
     print(response.content)
 """
 
-__version__ = "2.3.5"
+from .utils.version import __version__
 
 from .core.factory import create_llm
 from .core.session import BasicSession
@@ -39,7 +40,7 @@ try:
 except ImportError:
     _has_embeddings = False
 
-# Processing module (core functionality - always available)
+# Processing module (core functionality)
 from .processing import BasicSummarizer, SummaryStyle, SummaryLength, BasicExtractor
 _has_processing = True
 
@@ -59,5 +60,5 @@ __all__ = [
 if _has_embeddings:
     __all__.append('EmbeddingManager')
 
-# Processing is always available now
+# Processing is core functionality
 __all__.extend(['BasicSummarizer', 'SummaryStyle', 'SummaryLength', 'BasicExtractor'])

abstractllm/apps/__main__.py

@@ -8,10 +8,12 @@ Usage:
 Available apps:
     summarizer    - Document summarization tool
     extractor     - Entity and relationship extraction tool
+    judge         - Text evaluation and scoring tool
 
 Examples:
     python -m abstractllm.apps summarizer document.txt
     python -m abstractllm.apps extractor report.txt --format json-ld
+    python -m abstractllm.apps judge essay.txt --criteria clarity,accuracy
     python -m abstractllm.apps <app> --help
 """
 
@@ -36,9 +38,14 @@ def main():
         sys.argv = [sys.argv[0]] + sys.argv[2:]
         from .extractor import main as extractor_main
         extractor_main()
+    elif app_name == "judge":
+        # Remove the app name from sys.argv and run judge
+        sys.argv = [sys.argv[0]] + sys.argv[2:]
+        from .judge import main as judge_main
+        judge_main()
     else:
         print(f"Unknown app: {app_name}")
-        print("\nAvailable apps: summarizer, extractor")
+        print("\nAvailable apps: summarizer, extractor, judge")
         sys.exit(1)
 
 if __name__ == "__main__":

abstractllm/apps/extractor.py

@@ -6,31 +6,33 @@ Usage:
     python -m abstractllm.apps.extractor <file_path> [options]
 
 Options:
-    --focus=<focus>             Specific focus area for extraction (e.g., "technology", "business", "medical")
-    --style=<style>             Extraction style (structured, focused, minimal, comprehensive, default: structured)
-    --length=<length>           Extraction depth (brief, standard, detailed, comprehensive, default: standard)
-    --entity-types=<types>      Comma-separated entity types to focus on (person,organization,location,etc.)
-    --similarity-threshold=<t>  Similarity threshold for entity deduplication (0.0-1.0, default: 0.85)
-    --format=<format>          Output format (json-ld, json, yaml, default: json-ld)
-    --output=<output>          Output file path (optional, prints to console if not provided)
-    --chunk-size=<size>        Chunk size in characters (default: 6000, max: 32000)
-    --provider=<provider>      LLM provider (requires --model)
-    --model=<model>            LLM model (requires --provider)
-    --no-embeddings           Disable semantic entity deduplication
-    --fast                    Use fast extraction (skip verification, larger chunks, no embeddings)
-    --iterate=<number>        Number of refinement iterations (default: 1, finds missing entities and verifies relationships)
-    --minified                Output minified JSON-LD (compact, no indentation)
-    --verbose                 Show detailed progress information
-    --timeout=<seconds>       HTTP timeout for LLM providers (default: 300, increase for large models)
-    --help                    Show this help message
+    --focus <focus>             Specific focus area for extraction (e.g., "technology", "business", "medical")
+    --style <style>             Extraction style (structured, focused, minimal, comprehensive, default: structured)
+    --length <length>           Extraction depth (brief, standard, detailed, comprehensive, default: standard)
+    --entity-types <types>      Comma-separated entity types to focus on (person,organization,location,etc.)
+    --similarity-threshold <t>  Similarity threshold for entity deduplication (0.0-1.0, default: 0.85)
+    --format <format>           Output format (json-ld, json, yaml, triples, default: json-ld)
+    --output <output>           Output file path (optional, prints to console if not provided)
+    --chunk-size <size>         Chunk size in characters (default: 6000, max: 32000)
+    --provider <provider>       LLM provider (requires --model)
+    --model <model>             LLM model (requires --provider)
+    --no-embeddings             Disable semantic entity deduplication
+    --fast                      Use fast extraction (skip verification, larger chunks, no embeddings)
+    --iterate <number>          Number of refinement iterations (default: 1, finds missing entities and verifies relationships)
+    --minified                  Output minified JSON-LD (compact, no indentation)
+    --verbose                   Show detailed progress information
+    --timeout <seconds>         HTTP timeout for LLM providers (default: 300, increase for large models)
+    --max-tokens <tokens>       Maximum total tokens for LLM context (default: 32000)
+    --max-output-tokens <tokens> Maximum tokens for LLM output generation (default: 8000)
+    --help                      Show this help message
 
 Examples:
     python -m abstractllm.apps.extractor document.pdf
-    python -m abstractllm.apps.extractor report.txt --focus=technology --style=structured --verbose
-    python -m abstractllm.apps.extractor data.md --entity-types=person,organization --output=kg.jsonld
+    python -m abstractllm.apps.extractor report.txt --focus technology --style structured --verbose
+    python -m abstractllm.apps.extractor data.md --entity-types person,organization --output kg.jsonld
     python -m abstractllm.apps.extractor large.txt --fast --minified --verbose  # Fast, compact output
-    python -m abstractllm.apps.extractor report.txt --length=detailed --provider=openai --model=gpt-4o-mini
-    python -m abstractllm.apps.extractor doc.txt --iterate=3 --verbose  # 3 refinement passes for higher quality
+    python -m abstractllm.apps.extractor report.txt --length detailed --provider openai --model gpt-4o-mini
+    python -m abstractllm.apps.extractor doc.txt --iterate 3 --verbose  # 3 refinement passes for higher quality
 """
 
 import argparse
@@ -239,8 +241,8 @@ Default model setup:
     parser.add_argument(
         '--chunk-size',
         type=int,
-        default=6000,
-        help='Chunk size in characters (default: 6000, max: 32000)'
+        default=8000,
+        help='Chunk size in characters (default: 8000, max: 32000)'
     )
 
     parser.add_argument(
@@ -294,10 +296,24 @@ Default model setup:
     parser.add_argument(
         '--timeout',
         type=float,
-        default=300.0,
+        default=600.0,
         help='HTTP request timeout in seconds for LLM providers (default: 300, i.e., 5 minutes). Increase for large models like 36B+ parameters'
     )
 
+    parser.add_argument(
+        '--max-tokens',
+        type=int,
+        default=32000,
+        help='Maximum total tokens for LLM context (default: 32000)'
+    )
+
+    parser.add_argument(
+        '--max-output-tokens',
+        type=int,
+        default=8000,
+        help='Maximum tokens for LLM output generation (default: 8000)'
+    )
+
     # Parse arguments
     args = parser.parse_args()
 
@@ -397,24 +413,28 @@ Default model setup:
                 max_output_tokens = 8000
 
             if args.verbose:
-                print(f"Initializing BasicExtractor (mode: {extraction_mode}, {args.provider}, {args.model}, {max_tokens} token context, {max_output_tokens} output tokens)...")
+                print(f"Initializing BasicExtractor (mode: {extraction_mode}, {args.provider}, {args.model}, {args.max_tokens} token context, {args.max_output_tokens} output tokens)...")
                 if adjusted_chunk_size != args.chunk_size:
                     print(f"Adjusted chunk size from {args.chunk_size} to {adjusted_chunk_size} characters for {args.provider} compatibility")
 
-            llm = create_llm(args.provider, model=args.model, max_tokens=max_tokens, max_output_tokens=max_output_tokens, timeout=args.timeout)
+            llm = create_llm(args.provider, model=args.model, max_tokens=args.max_tokens, max_output_tokens=args.max_output_tokens, timeout=args.timeout)
 
             extractor = BasicExtractor(
                 llm=llm,
-                max_chunk_size=adjusted_chunk_size
+                max_chunk_size=adjusted_chunk_size,
+                max_tokens=args.max_tokens,
+                max_output_tokens=args.max_output_tokens
             )
         else:
             # Default configuration
             if args.verbose:
-                print(f"Initializing BasicExtractor (mode: {extraction_mode}, ollama, qwen3:4b-instruct-2507-q4_K_M, 32000 token context, 8000 output tokens)...")
+                print(f"Initializing BasicExtractor (mode: {extraction_mode}, ollama, qwen3:4b-instruct-2507-q4_K_M, {args.max_tokens} token context, {args.max_output_tokens} output tokens)...")
 
             try:
                 extractor = BasicExtractor(
-                    max_chunk_size=args.chunk_size
+                    max_chunk_size=args.chunk_size,
+                    max_tokens=args.max_tokens,
+                    max_output_tokens=args.max_output_tokens
                 )
             except RuntimeError as e:
                 # Handle default model not available

abstractllm/apps/judge.py

@@ -6,34 +6,37 @@ Usage:
     python -m abstractllm.apps.judge <file_path_or_text> [file2] [file3] ... [options]
 
 Options:
-    --context=<context>            Evaluation context description (e.g., "code review", "documentation assessment")
-    --criteria=<criteria>          Comma-separated criteria to evaluate (clarity,simplicity,actionability,soundness,innovation,effectiveness,relevance,completeness,coherence)
-    --custom-criteria=<criteria>   Comma-separated custom criteria (e.g., "has_examples,covers_edge_cases")
-    --reference=<file_or_text>     Reference content for comparison-based evaluation
-    --format=<format>             Output format (json, yaml, plain, default: json)
-    --output=<output>             Output file path (optional, prints to console if not provided)
-    --provider=<provider>         LLM provider (requires --model)
-    --model=<model>               LLM model (requires --provider)
-    --temperature=<temp>          Temperature for evaluation (default: 0.1 for consistency)
-    --verbose                     Show detailed progress information
-    --include-criteria           Include detailed explanation of evaluation criteria in assessment
-    --timeout=<seconds>           HTTP timeout for LLM providers (default: 300)
-    --help                       Show this help message
+    --context <context>             Evaluation context description (e.g., "code review", "documentation assessment")
+    --criteria <criteria>           Comma-separated criteria to evaluate (clarity,simplicity,actionability,soundness,innovation,effectiveness,relevance,completeness,coherence)
+    --focus <focus>                  Specific focus areas for evaluation (e.g., "technical accuracy,performance,security")
+    --reference <file_or_text>      Reference content for comparison-based evaluation
+    --format <format>               Output format (json, yaml, plain, default: json)
+    --output <output>               Output file path (optional, prints to console if not provided)
+    --provider <provider>           LLM provider (requires --model)
+    --model <model>                 LLM model (requires --provider)
+    --temperature <temp>            Temperature for evaluation (default: 0.1 for consistency)
+    --max-tokens <tokens>           Maximum total tokens for LLM context (default: 32000)
+    --max-output-tokens <tokens>    Maximum tokens for LLM output generation (default: 8000)
+    --verbose                       Show detailed progress information
+    --debug                         Show raw LLM responses and detailed debugging information
+    --include-criteria              Include detailed explanation of evaluation criteria in assessment
+    --timeout <seconds>             HTTP timeout for LLM providers (default: 300)
+    --help                          Show this help message
 
 Examples:
     # Single file or text
     python -m abstractllm.apps.judge "This code is well-structured and solves the problem efficiently."
-    python -m abstractllm.apps.judge document.py --context="code review" --criteria=clarity,soundness,effectiveness
+    python -m abstractllm.apps.judge document.py --context "code review" --criteria clarity,soundness,effectiveness
 
     # Multiple files (evaluated sequentially to avoid context overflow)
-    python -m abstractllm.apps.judge file1.py file2.py file3.py --context="code review" --output=assessments.json
-    python -m abstractllm.apps.judge *.py --context="Python code review" --format=plain
-    python -m abstractllm.apps.judge docs/*.md --context="documentation review" --criteria=clarity,completeness
+    python -m abstractllm.apps.judge file1.py file2.py file3.py --context "code review" --output assessments.json
+    python -m abstractllm.apps.judge *.py --context "Python code review" --format plain
+    python -m abstractllm.apps.judge docs/*.md --context "documentation review" --criteria clarity,completeness
 
     # Other options
-    python -m abstractllm.apps.judge proposal.md --custom-criteria=has_examples,addresses_concerns --output=assessment.json
-    python -m abstractllm.apps.judge content.txt --reference=ideal_solution.txt --format=plain --verbose
-    python -m abstractllm.apps.judge text.md --provider=openai --model=gpt-4o-mini --temperature=0.05
+    python -m abstractllm.apps.judge proposal.md --focus "technical accuracy,completeness,examples" --output assessment.json
+    python -m abstractllm.apps.judge content.txt --reference ideal_solution.txt --format plain --verbose
+    python -m abstractllm.apps.judge text.md --provider openai --model gpt-4o-mini --temperature 0.05
 """
 
 import argparse
@@ -253,16 +256,16 @@ def main():
 Examples:
   # Single file or text
   python -m abstractllm.apps.judge "This code is well-structured."
-  python -m abstractllm.apps.judge document.py --context="code review" --criteria=clarity,soundness
-  python -m abstractllm.apps.judge proposal.md --custom-criteria=has_examples --output=assessment.json
+  python -m abstractllm.apps.judge document.py --context "code review" --criteria clarity,soundness
+  python -m abstractllm.apps.judge proposal.md --focus "technical accuracy,examples" --output assessment.json
 
   # Multiple files (evaluated sequentially)
-  python -m abstractllm.apps.judge file1.py file2.py file3.py --context="code review" --format=json
-  python -m abstractllm.apps.judge docs/*.md --context="documentation review" --format=plain
+  python -m abstractllm.apps.judge file1.py file2.py file3.py --context "code review" --format json
+  python -m abstractllm.apps.judge docs/*.md --context "documentation review" --format plain
 
   # Other options
-  python -m abstractllm.apps.judge content.txt --reference=ideal.txt --format=plain --verbose
-  python -m abstractllm.apps.judge text.md --provider=openai --model=gpt-4o-mini
+  python -m abstractllm.apps.judge content.txt --reference ideal.txt --format plain --verbose
+  python -m abstractllm.apps.judge text.md --provider openai --model gpt-4o-mini
 
 Available criteria:
   clarity, simplicity, actionability, soundness, innovation, effectiveness,
@@ -297,8 +300,8 @@ Default model setup:
     )
 
     parser.add_argument(
-        '--custom-criteria',
-        help='Comma-separated custom criteria (e.g., "has_examples,covers_edge_cases")'
+        '--focus',
+        help='Specific focus areas for evaluation (e.g., "technical accuracy,performance,security")'
     )
 
     parser.add_argument(
@@ -340,12 +343,32 @@ Default model setup:
         help='Temperature for evaluation (default: 0.1 for consistency)'
     )
 
+    parser.add_argument(
+        '--max-tokens',
+        type=int,
+        default=32000,
+        help='Maximum total tokens for LLM context (default: 32000)'
+    )
+
+    parser.add_argument(
+        '--max-output-tokens',
+        type=int,
+        default=8000,
+        help='Maximum tokens for LLM output generation (default: 8000)'
+    )
+
     parser.add_argument(
         '--verbose',
         action='store_true',
         help='Show detailed progress information'
     )
 
+    parser.add_argument(
+        '--debug',
+        action='store_true',
+        help='Show raw LLM responses and detailed debugging information'
+    )
+
     parser.add_argument(
         '--include-criteria',
         action='store_true',
@@ -421,26 +444,34 @@ Default model setup:
 
         # Parse criteria
         criteria_list = parse_criteria_list(args.criteria)
-        custom_criteria = parse_criteria_list(args.custom_criteria)
+        focus = args.focus
         judgment_criteria = build_judgment_criteria(criteria_list)
 
         # Initialize judge
         if args.provider and args.model:
             if args.verbose:
-                print(f"Initializing BasicJudge ({args.provider}, {args.model}, temperature={args.temperature})...")
+                print(f"Initializing BasicJudge ({args.provider}, {args.model}, temperature={args.temperature}, {args.max_tokens} token context, {args.max_output_tokens} output tokens)...")
 
             judge = create_judge(
                 provider=args.provider,
                 model=args.model,
                 temperature=args.temperature,
+                max_tokens=args.max_tokens,
+                max_output_tokens=args.max_output_tokens,
+                debug=args.debug,
                 timeout=args.timeout
             )
         else:
             if args.verbose:
-                print(f"Initializing BasicJudge (ollama, qwen3:4b-instruct-2507-q4_K_M, temperature={args.temperature})...")
+                print(f"Initializing BasicJudge (ollama, qwen3:4b-instruct-2507-q4_K_M, temperature={args.temperature}, {args.max_tokens} token context, {args.max_output_tokens} output tokens)...")
 
             try:
-                judge = BasicJudge(temperature=args.temperature)
+                judge = BasicJudge(
+                    temperature=args.temperature,
+                    max_tokens=args.max_tokens,
+                    max_output_tokens=args.max_output_tokens,
+                    debug=args.debug
+                )
             except RuntimeError as e:
                 print(f"\n{e}")
                 print("\n🚀 Quick alternatives to get started:")
@@ -467,7 +498,7 @@ Default model setup:
                     file_paths=single_input,
                     context=args.context,
                     criteria=judgment_criteria,
-                    custom_criteria=custom_criteria,
+                    focus=focus,
                     reference=reference,
                     include_criteria=args.include_criteria,
                     exclude_global=args.exclude_global
@@ -478,7 +509,7 @@ Default model setup:
                     content=single_input,
                     context=args.context,
                     criteria=judgment_criteria,
-                    custom_criteria=custom_criteria,
+                    focus=focus,
                     reference=reference,
                     include_criteria=args.include_criteria
                 )
@@ -488,7 +519,7 @@ Default model setup:
                 file_paths=file_or_text,
                 context=args.context,
                 criteria=judgment_criteria,
-                custom_criteria=custom_criteria,
+                focus=focus,
                 reference=reference,
                 include_criteria=args.include_criteria,
                 exclude_global=args.exclude_global