gitpr-cli 1.7.0__py3-none-any.whl → 1.8.0__py3-none-any.whl

{gitpr_cli-1.7.0.dist-info → gitpr_cli-1.8.0.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: gitpr-cli
-Version: 1.7.0
+Version: 1.8.0
 Summary: AI-powered PR automation, commit messages, and code review (Gemini, DeepSeek, and Ollama)
 Author-email: Natan Fiuza <contato@natanfiuza.dev.br>
 Requires-Python: >=3.10
@@ -499,7 +499,7 @@ This is useful for scripting, CI/CD pipelines, and one-off queries where you don
 
 📖 **Full documentation:** [gitpr.natanfiuza.dev.br/docs/mcp-integration](https://gitpr.natanfiuza.dev.br/docs/mcp-integration) — available in 5 languages (EN, PT-BR, PT-PT, ES, FR).
 
-> 🧩 **Official VS Code & Cursor Extension** — Prefer native editor commands, SCM toolbar buttons, and inline `Problems` diagnostics? Download the ready-to-use `.vsix` from [GitHub Releases](https://github.com/natanfiuza/gitpr/releases) or see the [Editor Extension Guide](https://gitpr.natanfiuza.dev.br/docs/editor-extension) for 1-click install in VS Code and Cursor.
+> 🧩 **Official VS Code & Cursor Extension** — Prefer native editor commands, SCM toolbar buttons, and inline `Problems` diagnostics? Download the ready-to-use `.vsix` from [GitHub Releases](https://github.com/natanfiuza/gitpr/releases) or see the [Editor Extension Guide](https://gitpr.natanfiuza.dev.br/docs/editor-extension) for 1-click install in VS Code and Cursor. Maintainers can run `scripts/marketplace-setup-wizard.sh` to configure and publish to Visual Studio Marketplace and Open VSX Registry.
 
 > 💬 **MCP Prompts** — GitPR also exposes 7 pre-defined message templates (prompts) for common flows like "Review PR", "Generate Commit Message", and "Create Issue from Diff". See the [MCP Prompts guide](https://gitpr.natanfiuza.dev.br/docs/mcp-prompts) for the full list.
 

{gitpr_cli-1.7.0.dist-info → gitpr_cli-1.8.0.dist-info}/RECORD

@@ -1,4 +1,4 @@
-gitpr_cli-1.7.0.dist-info/licenses/LICENSE,sha256=UQzLC6mwqb-ZNDkBBAqOxC-SMQcEID0gaM6oFcu6u48,27030
+gitpr_cli-1.8.0.dist-info/licenses/LICENSE,sha256=UQzLC6mwqb-ZNDkBBAqOxC-SMQcEID0gaM6oFcu6u48,27030
 src/__init__.py,sha256=ZHymXsocMDazI_fjOIgAQsRQvFPN2tV8FOuv0qptI_o,30
 src/ai_providers.py,sha256=QyFEOsXyA59lZziO5bl-4nlYFJpxBcxU9aLhcTBnkWE,14633
 src/ascii_chart.py,sha256=WFQclaxm5W6bPBBNVlImyrnrJYCSj5gRgEpldYzI7wk,8279
@@ -7,14 +7,14 @@ src/cache.py,sha256=frROc69NyTn3TdJIOPctvuH9z_sh2pdKxL4ekMtyh6k,6572
 src/changelog_builder.py,sha256=Ea0coS6LCqsdxTWrKW8SZyVsD02HuNNYzCoggI0loi8,9922
 src/chat_memory.py,sha256=PcohnyspAfIwwwbsp9U2Wk_1AylJMFEMoWqRSJe_w30,8949
 src/commit_classifier.py,sha256=HQHxPbCfgW_GD1_4mbO2No16jh_Ci5RTSoLbTU4IWYI,6920
-src/config.py,sha256=MrfZW27x-41gI9P8gtAyMuTryzeluBtvZfxRZkVIiBk,47122
-src/config_schema.py,sha256=VATmOHWs_eeMLVhrtGFANcOVdrYwSadiFPUUhZM33OE,58091
-src/core.py,sha256=RmwqnEa0-lEHKpxCIH95JLJkJQGsfx_QC8n8H_uJitQ,99037
+src/config.py,sha256=1U9dZyfk_OsMzoIcYIbvh_N1H9ZO34PikbjjCHzg7PI,48895
+src/config_schema.py,sha256=qClFT_qoa_oV3yyPENiT79MB4AFsuXvsSlo4an-DIcE,60375
+src/core.py,sha256=9IuJPJecTyATUzI4qCXWv67PnfzQMulrKOOb24CdN3U,106050
 src/diff_parser.py,sha256=JHeaTJflNOEbR9SYK_Ljr2hctrrABUIb31IFHAbyIDo,15405
 src/doc_links.py,sha256=GdANhE6XbEnVDsbpCf22eUZnqFXor_2SU1O4S-pteIE,2028
 src/github_api.py,sha256=i1o58oKWWmMCerP-g-hOQHGCwPWd01ooKntTKjpFhIA,3948
 src/i18n.py,sha256=Xp9c1dNGmTHbo8SyXa-LQuO71nWx3RP-Us7jJ7ARtUY,4804
-src/issue_engine.py,sha256=RmZsQVg6N1vCcpMdv_5FX-cMGgQ3I0vwMN4parQ4zFg,10974
+src/issue_engine.py,sha256=fgf1SO0ENJmHTKP4NFlHaT2jdOWgWAVdU0yRoQq2-hc,11171
 src/ledger.py,sha256=3p5nX9gPGHgANQxBxEywrKo3TNb5KdWDeTI2mslYopo,43787
 src/linter_engine.py,sha256=hIxht8M09Fnm6NiJZOqjI6dyqnFoHUDVA8Ve4dWZV4s,25214
 src/linter_wizard.py,sha256=mo8aSTZmVLluCOskWv3nji74WPJh_6c122zwADN9aCc,7634
@@ -26,14 +26,14 @@ src/project_config.py,sha256=r-gzjKeFuTYvUzOflQlpbijn6uLXtPEUBzA-1PzUR_k,16749
 src/release_engine.py,sha256=S0kgO-a7RSC07AzvRWRlGo32jx38PAQGIzgvYXOfthk,41274
 src/reviewer_resolution.py,sha256=bFRdf4GVWIpz4kk7cyF7EYWV43aP7V8sAlNGqKRPd7A,5913
 src/reviewer_suggestion.py,sha256=YeXJLjJjn1YS69FChWZBF4qHUuIwvv_n-GPiymLN8LQ,8904
-src/scm_settings.py,sha256=3pXgBbxqekvD4LM-I3GgfV4IOa6QJn4x3-DPAThCris,14156
+src/scm_settings.py,sha256=ZNMgr893qj-B3DjbXkuINGnJYvJU_WvHvi535ZjdVow,14698
 src/security.py,sha256=fkWy-r4tdISH0xOyWZyR8ZCFyLuYqu8fn9sKsIeNAso,1351
 src/security_ruleset.py,sha256=4mqmqtWLHuhuCcjjaoKUM_YFzil4cWZ_ubjlOLgbM_4,4874
 src/skill_frontmatter.py,sha256=K-HsjwSYPHUsDRjIiugBE3koxVdGf38L5d4l8eYVqEo,2167
 src/spinner.py,sha256=JGOO_taTXbrQK0KnNxUQ40PcpS802bYYBosWWm_NG9o,9012
 src/suggest_reviewers.py,sha256=LTOtwDIjaBS_FGDU6bjN6oGy-YD112hIPugYytNKD_A,5821
 src/tui_issue.py,sha256=yaNpj9SSXiWWjMJ7a9FABcNWSffCkA1tjUSnJ1hN-MQ,11477
-src/updater.py,sha256=8OwxL8hhh-hzKyQo5sjckw_e4bNlmlKeegwwH35nrG8,9560
+src/updater.py,sha256=ECdRPKQ-_sKDi4LFn2MlMXOjTPFx7QdJApjmv5pY_lE,9560
 src/usage_log.py,sha256=CBCNlUm_xftXsBaGRHvsEjTENvZfE-lW70HGLMkisSg,4508
 src/version_bump.py,sha256=Dw1LZ_c6VzaO_s9cDeYSevoWtXVr_oP70eS_mkrOm8s,3041
 src/application/__init__.py,sha256=3YXFcmxJh0f7BwryKXxUrOhGrLnZLsagV-VbBIk0eGk,31
@@ -128,7 +128,7 @@ src/infrastructure/scm/__init__.py,sha256=CbW2p9DsYpwwUqy4xdtgoPugsLijopfNwLT-IA
 src/infrastructure/scm/azure_devops_provider.py,sha256=9IpgB-Vqzgj0VDFXaYs4PFRiecsQGU9svW5m3OpefsI,18790
 src/infrastructure/scm/base.py,sha256=9tN2IivEngHoo1pDK3PivWHvHSkXokuyQQUJ680gsXU,17198
 src/infrastructure/scm/bitbucket_provider.py,sha256=A58dMDa60lLUynENrinWENFz5dSA_lb9VJSg22ptbkI,16895
-src/infrastructure/scm/factory.py,sha256=O3bRT6f15AozcldZfiH2LgOQ4g4GqMb0sbEWCIeF0HE,4487
+src/infrastructure/scm/factory.py,sha256=olcDDs0tsa8Ae0vKhhW5DnS408xRhn5fxv_0DskFCgY,4665
 src/infrastructure/scm/github_provider.py,sha256=vdJDTTYd-CYkAQos-6IZ2hFOJkXG09SrkIcZzHbuuIc,21316
 src/infrastructure/scm/gitlab_provider.py,sha256=VAYjdg5gU43FDamXUNMdFzs9MwIGyNRqMzEnLTacRE8,17951
 src/infrastructure/scm/web_links.py,sha256=-7oF7oUwK_UvKVJZWDPbTnKBi9JEQu2dMDYtgqH10iM,3724
@@ -153,7 +153,7 @@ src/policy_packs/vue-quality/policy.yml,sha256=FyVL17Fl_bgVjxZU5lO7MME5w04vjKV3U
 src/review/__init__.py,sha256=ux2rkwebfnzuQpFc57eEceOC4CZkojPOtfp951Ux5Gc,700
 src/review/diff_normalizer.py,sha256=8KnFbauGK-fuLGSSL_ox4kAwCE048MuTmorPwBrhlEU,3675
 src/review/diff_source.py,sha256=V6fMtgxNJbMBL_qRifkLGfzFwtaTCbFFKFvcutHZjLk,2733
-src/review/remote_pr.py,sha256=9oEEo-v-D08egxwN_EINh5Ju74vow91-CAH1kAb2ql0,13669
+src/review/remote_pr.py,sha256=f9gkvK4uw--Np_la4lRuSkhmEplUccI5ePCllJFj-Lo,13884
 src/review/render.py,sha256=l63vQ3WiGxPSzCk4JxyQoFNfIyG0AdHgMobAjMeaOcY,3194
 src/split/__init__.py,sha256=8XZn4jwPhQ0pJyNEMTBfgbGTMJkQUvePNp5ZncWCJ-Q,929
 src/split/apply_split_plan.py,sha256=ZKoACWLCaBndLzvRl8wHyAyPlalTrjrPJ2ep_AjpWCE,10929
@@ -177,8 +177,8 @@ src/ui/demo/__init__.py,sha256=aOdUN0izuwW2FfLAHG9Qga-DJG5NBVqP0gYNO8OiKls,59
 src/ui/demo/demo_app.py,sha256=4t_Mct-BJXpe5tlt1XmtRyN1IT0OL6U3WSmGqOJqMG4,3165
 src/ui/demo/demo_help_screen.py,sha256=I_LELswhPtbHjfMWGVUqLtOwxGbzSYJsVWXjqZrLoLA,2108
 src/ui/demo/demo_screens.py,sha256=GXaQFTgTn9awSP-EYiBXsCZqB583WrSQcuNYoBK12g8,1997
-gitpr_cli-1.7.0.dist-info/METADATA,sha256=hpoCCrRp4v_yMf_v7Pqsg4YtA0nVWFha4N2JugWwr7E,54113
-gitpr_cli-1.7.0.dist-info/WHEEL,sha256=mOdXR9C3wMVov7cSY8PY-XCzo3SmmWCHvGQr5csJoao,91
-gitpr_cli-1.7.0.dist-info/entry_points.txt,sha256=TstGnMUzdOI5qNmpM97EPDBPuMqOZxoo6EoLxHSOMSQ,71
-gitpr_cli-1.7.0.dist-info/top_level.txt,sha256=74rtVfumQlgAPzR5_2CgYN24MB0XARCg0t-gzk6gTrM,4
-gitpr_cli-1.7.0.dist-info/RECORD,,
+gitpr_cli-1.8.0.dist-info/METADATA,sha256=6tgDZ4ZC09Mkw_S35xuA2v_ciLAMqFoM7b5buvBxdNM,54248
+gitpr_cli-1.8.0.dist-info/WHEEL,sha256=mOdXR9C3wMVov7cSY8PY-XCzo3SmmWCHvGQr5csJoao,91
+gitpr_cli-1.8.0.dist-info/entry_points.txt,sha256=TstGnMUzdOI5qNmpM97EPDBPuMqOZxoo6EoLxHSOMSQ,71
+gitpr_cli-1.8.0.dist-info/top_level.txt,sha256=74rtVfumQlgAPzR5_2CgYN24MB0XARCg0t-gzk6gTrM,4
+gitpr_cli-1.8.0.dist-info/RECORD,,

src/config.py

@@ -249,6 +249,59 @@ def get_linter_timeout():
     return _positive_float(os.getenv("GITPR_LINTER_TIMEOUT"), _DEFAULT_LINTER_TIMEOUT)
 
 
+DEFAULT_MAP_REDUCE_MAX_TOKENS = 90000
+DEFAULT_MAP_REDUCE_MAX_PART_TOKENS = 50000
+
+
+def _positive_int_from_raw(raw, fallback):
+    """Parses raw as a positive int, falling back on non-positive or invalid input."""
+    if raw is None:
+        return fallback
+    try:
+        val = int(str(raw).strip())
+        return val if val > 0 else fallback
+    except (TypeError, ValueError):
+        return fallback
+
+
+def get_map_reduce_thresholds(provider=None):
+    """Returns (max_tokens, max_part_tokens) for Map-Reduce chunking.
+
+    Resolution hierarchy:
+    1. Provider-specific: GITPR_MAP_REDUCE_MAX_TOKENS_<PROVIDER>
+    2. Generic: GITPR_MAP_REDUCE_MAX_TOKENS
+    3. Baseline default: 90,000 tokens
+
+    For part tokens:
+    1. Provider-specific: GITPR_MAP_REDUCE_MAX_PART_TOKENS_<PROVIDER>
+    2. Generic: GITPR_MAP_REDUCE_MAX_PART_TOKENS
+    3. Baseline default: 50,000 tokens
+
+    Guarantees max_part_tokens <= max_tokens.
+    """
+    load_dotenv(ENV_FILE)
+    prov_key = (provider or "").strip().upper()
+
+    trigger_val = None
+    if prov_key:
+        trigger_val = os.getenv(f"GITPR_MAP_REDUCE_MAX_TOKENS_{prov_key}")
+    if trigger_val is None or not str(trigger_val).strip():
+        trigger_val = os.getenv("GITPR_MAP_REDUCE_MAX_TOKENS")
+    max_tokens = _positive_int_from_raw(trigger_val, DEFAULT_MAP_REDUCE_MAX_TOKENS)
+
+    part_val = None
+    if prov_key:
+        part_val = os.getenv(f"GITPR_MAP_REDUCE_MAX_PART_TOKENS_{prov_key}")
+    if part_val is None or not str(part_val).strip():
+        part_val = os.getenv("GITPR_MAP_REDUCE_MAX_PART_TOKENS")
+    max_part_tokens = _positive_int_from_raw(part_val, DEFAULT_MAP_REDUCE_MAX_PART_TOKENS)
+
+    if max_part_tokens > max_tokens:
+        max_part_tokens = max_tokens
+
+    return max_tokens, max_part_tokens
+
+
 # Metrics pricing.
 #
 # Two layers, the second overriding the first per model:

src/config_schema.py

@@ -483,6 +483,28 @@ FIELDS = (
         validator=VALIDATOR_AI_KEY,
         show_if=(("DEFAULT_AI_PROVIDER", ("gemini",)),),
     ),
+    ConfigField(
+        key="GITPR_MAP_REDUCE_MAX_TOKENS_GEMINI",
+        label=__("Gemini Map-Reduce Trigger Tokens"),
+        description=__(
+            "Maximum total diff tokens before triggering Map-Reduce processing for Gemini."
+        ),
+        category="ai",
+        kind=KIND_INT,
+        default="90000",
+        show_if=(("DEFAULT_AI_PROVIDER", ("gemini",)),),
+    ),
+    ConfigField(
+        key="GITPR_MAP_REDUCE_MAX_PART_TOKENS_GEMINI",
+        label=__("Gemini Map-Reduce Part Size"),
+        description=__(
+            "Maximum tokens per batch chunk during Gemini Map-Reduce processing."
+        ),
+        category="ai",
+        kind=KIND_INT,
+        default="50000",
+        show_if=(("DEFAULT_AI_PROVIDER", ("gemini",)),),
+    ),
     ConfigField(
         key="DEEPSEEK_API_MODEL_PRIMARY",
         label=__("DeepSeek Advanced Model"),
@@ -559,6 +581,28 @@ FIELDS = (
         validator=VALIDATOR_AI_KEY,
         show_if=(("DEFAULT_AI_PROVIDER", ("deepseek",)),),
     ),
+    ConfigField(
+        key="GITPR_MAP_REDUCE_MAX_TOKENS_DEEPSEEK",
+        label=__("DeepSeek Map-Reduce Trigger Tokens"),
+        description=__(
+            "Maximum total diff tokens before triggering Map-Reduce processing for DeepSeek."
+        ),
+        category="ai",
+        kind=KIND_INT,
+        default="90000",
+        show_if=(("DEFAULT_AI_PROVIDER", ("deepseek",)),),
+    ),
+    ConfigField(
+        key="GITPR_MAP_REDUCE_MAX_PART_TOKENS_DEEPSEEK",
+        label=__("DeepSeek Map-Reduce Part Size"),
+        description=__(
+            "Maximum tokens per batch chunk during DeepSeek Map-Reduce processing."
+        ),
+        category="ai",
+        kind=KIND_INT,
+        default="50000",
+        show_if=(("DEFAULT_AI_PROVIDER", ("deepseek",)),),
+    ),
     ConfigField(
         key="OLLAMA_API_MODEL_PRIMARY",
         label=__("Ollama Advanced Model"),
@@ -579,6 +623,28 @@ FIELDS = (
         show_if=(("DEFAULT_AI_PROVIDER", ("ollama",)),),
         seeded=True,
     ),
+    ConfigField(
+        key="GITPR_MAP_REDUCE_MAX_TOKENS_OLLAMA",
+        label=__("Ollama Map-Reduce Trigger Tokens"),
+        description=__(
+            "Maximum total diff tokens before triggering Map-Reduce processing for Ollama."
+        ),
+        category="ai",
+        kind=KIND_INT,
+        default="90000",
+        show_if=(("DEFAULT_AI_PROVIDER", ("ollama",)),),
+    ),
+    ConfigField(
+        key="GITPR_MAP_REDUCE_MAX_PART_TOKENS_OLLAMA",
+        label=__("Ollama Map-Reduce Part Size"),
+        description=__(
+            "Maximum tokens per batch chunk during Ollama Map-Reduce processing."
+        ),
+        category="ai",
+        kind=KIND_INT,
+        default="50000",
+        show_if=(("DEFAULT_AI_PROVIDER", ("ollama",)),),
+    ),
     # ----------------------------------------------------------- Pull request
     ConfigField(
         key="OUTPUT_FILE_NAME",

src/core.py

@@ -1099,20 +1099,136 @@ def estimate_token_count(text):
     return len(text) // 4
 
 
-def split_diff_into_chunks(diff_text, max_tokens=90000):
-    """Splits the diff based on the token limit while preserving file header integrity."""
+def _split_line_chunk(text, max_tokens):
+    """Splits arbitrary text by lines (and character slices if needed) keeping each part <= max_tokens."""
+    chunks = []
+    current = ""
+    for line in text.splitlines(keepends=True):
+        if estimate_token_count(line) > max_tokens:
+            if current:
+                chunks.append(current)
+                current = ""
+            step = max_tokens * 4
+            for start in range(0, len(line), step):
+                slice_str = line[start : start + step]
+                chunks.append(slice_str)
+        elif estimate_token_count(current + line) > max_tokens and current:
+            chunks.append(current)
+            current = line
+        else:
+            current += line
+    if current:
+        chunks.append(current)
+    return chunks or [text]
+
+
+def _split_giant_file(file_diff, max_part_tokens):
+    """Splits a single file diff that exceeds max_part_tokens into smaller parts."""
+    fn_match = re.search(r"^diff --git a/(\S+) b/\S+", file_diff, flags=re.MULTILINE)
+    file_name = fn_match.group(1) if fn_match else "file"
+
+    hunk_match = re.search(
+        r"^@@ -\d+(?:,\d+)? \+\d+(?:,\d+)? @@", file_diff, flags=re.MULTILINE
+    )
+    if hunk_match:
+        header = file_diff[: hunk_match.start()]
+        body = file_diff[hunk_match.start() :]
+    else:
+        header = ""
+        body = file_diff
+
+    hunk_parts = re.split(
+        r"(^@@ -\d+(?:,\d+)? \+\d+(?:,\d+)? @@.*)", body, flags=re.MULTILINE
+    )
+    raw_hunks = []
+    if len(hunk_parts) > 1:
+        if hunk_parts[0].strip():
+            raw_hunks.append(hunk_parts[0])
+        for j in range(1, len(hunk_parts), 2):
+            hunk_content = hunk_parts[j] + (
+                hunk_parts[j + 1] if j + 1 < len(hunk_parts) else ""
+            )
+            raw_hunks.append(hunk_content)
+    else:
+        raw_hunks = [body]
+
+    sub_chunks = []
+    current_sub = ""
+    header_clean = header.strip()
+
+    def _format_part(sub_content, idx):
+        prefix = (
+            f"{header_clean}\n# [{file_name} - part {idx}]\n\n"
+            if header_clean
+            else f"# [{file_name} - part {idx}]\n\n"
+        )
+        return prefix + sub_content
+
+    part_idx = 1
+    for hunk in raw_hunks:
+        if estimate_token_count(hunk) > max_part_tokens:
+            if current_sub:
+                sub_chunks.append(_format_part(current_sub, part_idx))
+                part_idx += 1
+                current_sub = ""
+            line_pieces = _split_line_chunk(hunk, max_part_tokens)
+            for piece in line_pieces:
+                sub_chunks.append(_format_part(piece, part_idx))
+                part_idx += 1
+        elif (
+            estimate_token_count(current_sub + hunk) > max_part_tokens
+            and current_sub
+        ):
+            sub_chunks.append(_format_part(current_sub, part_idx))
+            part_idx += 1
+            current_sub = hunk
+        else:
+            current_sub += hunk
+
+    if current_sub:
+        sub_chunks.append(_format_part(current_sub, part_idx))
+
+    return sub_chunks or [file_diff]
+
+
+def split_diff_into_chunks(diff_text, max_tokens=90000, max_part_tokens=None):
+    """Splits the diff based on the token limit while preserving file header integrity.
+
+    - max_tokens: Diff size threshold that triggers Map-Reduce (default 90,000).
+    - max_part_tokens: Maximum tokens per batch chunk (default 50,000).
+    """
+    if max_part_tokens is None:
+        max_part_tokens = 50000
+    if max_part_tokens > max_tokens:
+        max_part_tokens = max_tokens
+
     if estimate_token_count(diff_text) <= max_tokens:
         return [diff_text]
 
-    parts = re.split(r"(^diff --git a/)", diff_text, flags=re.MULTILINE)
-    chunks = []
-    current_chunk = ""
+    parts = re.split(r"(^diff --git (?:[abciwos12]/|\S+ ))", diff_text, flags=re.MULTILINE)
+    if len(parts) <= 1:
+        return _split_line_chunk(diff_text, max_part_tokens)
 
+    file_diffs = []
+    if parts[0].strip():
+        file_diffs.append(parts[0])
     for i in range(1, len(parts), 2):
-        file_diff = parts[i] + parts[i + 1] if i + 1 < len(parts) else parts[i]
+        file_d = parts[i] + (parts[i + 1] if i + 1 < len(parts) else "")
+        file_diffs.append(file_d)
+
+    chunks = []
+    current_chunk = ""
 
-        if (
-            estimate_token_count(current_chunk + file_diff) > max_tokens
+    for file_diff in file_diffs:
+        file_tokens = estimate_token_count(file_diff)
+        if file_tokens > max_part_tokens:
+            if current_chunk:
+                chunks.append(current_chunk)
+                current_chunk = ""
+            file_sub_chunks = _split_giant_file(file_diff, max_part_tokens)
+            chunks.extend(file_sub_chunks)
+        elif (
+            estimate_token_count(current_chunk + file_diff) > max_part_tokens
             and current_chunk
         ):
             chunks.append(current_chunk)
@@ -1369,7 +1485,11 @@ def generate_pr_content(
         fg="cyan",
     )
 
-    chunks = split_diff_into_chunks(diff_text, max_tokens=90000)
+    from src.config import get_map_reduce_thresholds
+    max_tokens, max_part_tokens = get_map_reduce_thresholds(provider)
+    chunks = split_diff_into_chunks(
+        diff_text, max_tokens=max_tokens, max_part_tokens=max_part_tokens
+    )
     total_meta = {
         "prompt_tokens": 0,
         "completion_tokens": 0,
@@ -1479,6 +1599,70 @@ def generate_pr_content(
         )
         diff_unificado = "\n\n".join(resumos_parciais)
 
+        if estimate_token_count(diff_unificado) > max_tokens:
+            click.secho(
+                __("🔄 Large summary set detected! Running intermediate reduction..."),
+                fg="cyan",
+                dim=True,
+            )
+            intermediate_summaries = []
+            curr_batch = []
+            curr_tokens = 0
+            for item in resumos_parciais:
+                item_tok = estimate_token_count(item)
+                if curr_tokens + item_tok > max_part_tokens and curr_batch:
+                    batch_str = "\n\n".join(curr_batch)
+                    inter_prompt = (
+                        __(
+                            "Generate ONLY a JSON object in the format {json_format} consolidating these technical summaries into a unified summary:\n",
+                            json_format='{"resumo": "..."}',
+                        )
+                        + batch_str
+                    )
+                    inter_res = call_ai_model(
+                        provider,
+                        api_key,
+                        api_model,
+                        inter_prompt,
+                        instrucao_sistema,
+                        quiet=True,
+                        action=f"{action_folder}_inter_reduce",
+                    )
+                    if inter_res and "resumo" in inter_res:
+                        intermediate_summaries.append(inter_res["resumo"])
+                        _aggregate_meta(inter_res.pop("_telemetry_meta", None))
+                    else:
+                        intermediate_summaries.append(batch_str)
+                    curr_batch = [item]
+                    curr_tokens = item_tok
+                else:
+                    curr_batch.append(item)
+                    curr_tokens += item_tok
+            if curr_batch:
+                batch_str = "\n\n".join(curr_batch)
+                inter_prompt = (
+                    __(
+                        "Generate ONLY a JSON object in the format {json_format} consolidating these technical summaries into a unified summary:\n",
+                        json_format='{"resumo": "..."}',
+                    )
+                    + batch_str
+                )
+                inter_res = call_ai_model(
+                    provider,
+                    api_key,
+                    api_model,
+                    inter_prompt,
+                    instrucao_sistema,
+                    quiet=True,
+                    action=f"{action_folder}_inter_reduce",
+                )
+                if inter_res and "resumo" in inter_res:
+                    intermediate_summaries.append(inter_res["resumo"])
+                    _aggregate_meta(inter_res.pop("_telemetry_meta", None))
+                else:
+                    intermediate_summaries.append(batch_str)
+            diff_unificado = "\n\n".join(intermediate_summaries)
+
         if action_type == "commit":
             prompt = (
                 __(

src/infrastructure/scm/factory.py

@@ -70,10 +70,13 @@ def resolve_scm_provider(config: dict) -> ScmProvider:
 def detect_provider_from_remote(remote_url: str, default: Optional[str] = "github") -> Optional[str]:
     """Guess the forge from a git remote URL (case-insensitive substrings).
 
-    Order matters only for URLs that match several hosts — none exist today.
-    Anything unknown returns *default*: "github" reproduces the historical
-    behavior, while ``default=None`` lets a caller tell "recognised as GitHub"
-    apart from "not recognised at all" (self-hosted hosts need the latter).
+    The named forges are checked first, so a URL that mentions one of them wins
+    even when it also contains "github". Anything unknown returns *default*:
+    "github" reproduces the historical behavior, while ``default=None`` lets a
+    caller tell "recognised as GitHub" apart from "not recognised at all"
+    (self-hosted hosts need the latter). For that reason the GitHub check runs
+    before the ``None`` return: ``github.com`` is a recognised forge, not an
+    unknown host.
     """
     url = (remote_url or "").lower()
     if "gitlab" in url:
@@ -82,10 +85,10 @@ def detect_provider_from_remote(remote_url: str, default: Optional[str] = "githu
         return "bitbucket"
     if "dev.azure.com" in url or "visualstudio.com" in url:
         return "azure_devops"
-    if default is None:
-        return None
     if "github" in url:
         return "github"
+    if default is None:
+        return None
     return default
 
 

src/issue_engine.py

@@ -170,7 +170,12 @@ def generate_issue_content(context_text, context_type="diff"):
             total_meta["duration_ms"] += new_meta.get("duration_ms", 0)
 
     if context_type == "diff":
-        chunks = split_diff_into_chunks(context_text, max_tokens=90000)
+        from src.config import get_map_reduce_thresholds
+
+        max_tokens, max_part_tokens = get_map_reduce_thresholds(provider)
+        chunks = split_diff_into_chunks(
+            context_text, max_tokens=max_tokens, max_part_tokens=max_part_tokens
+        )
     else:
         chunks = [context_text]
 

src/review/remote_pr.py

@@ -342,9 +342,15 @@ def _chunk_count(diff_text):
     back, and the callers that need it (the MCP tool especially) have no
     terminal to read the "processing in N batches" line from.
     """
+    from src.config import get_map_reduce_thresholds
     from src.core import split_diff_into_chunks
 
     try:
-        return len(split_diff_into_chunks(diff_text, max_tokens=90000))
+        max_tokens, max_part_tokens = get_map_reduce_thresholds()
+        return len(
+            split_diff_into_chunks(
+                diff_text, max_tokens=max_tokens, max_part_tokens=max_part_tokens
+            )
+        )
     except Exception:
         return 1

src/scm_settings.py

@@ -366,6 +366,13 @@ def _ignored_provider_warning(choice) -> str:
     )
 
 
+def _default_provider_warning() -> str:
+    return __(
+        "⚠️ The origin remote does not name a forge, so GitPR is using GitHub. "
+        "Run 'gitpr --init' to choose the right forge for this repository."
+    )
+
+
 def resolve_scm_settings(
     start_dir=None, remote_url=None, config_path=None, warnings=None
 ) -> ScmSettings:
@@ -390,6 +397,11 @@ def resolve_scm_settings(
     choice = resolve_provider_choice(project_file.provider, remote)
     if choice.ignored:
         collected.append(_ignored_provider_warning(choice))
+    if remote and choice.source == SOURCE_DEFAULT:
+        # An existing remote that names no forge: the repository is about to be
+        # driven as GitHub unless someone records an override. The absent-remote
+        # case has its own error at the call sites.
+        collected.append(_default_provider_warning())
 
     token = _token_for(choice.key)
     if os.getenv(RAW_TOKEN_KEY):

src/updater.py

@@ -9,8 +9,8 @@ from packaging.version import InvalidVersion, Version
 # Canonical Single Source of Truth (SSOT) for GitPR package version (PyPI distribution)
 # ── Must stay ABOVE the i18n import: i18n.py lazily imports __lang_version__
 #     from here, and pyproject.toml reads __version__ via setuptools attr:.
-__version__ = "1.7.0"  # GitPR current package version (Canonical SSOT)
-__lang_version__ = "v0.0.42"  # Language dictionary version control (asset, independent of package version)
+__version__ = "1.8.0"  # GitPR current package version (Canonical SSOT)
+__lang_version__ = "v0.0.43"  # Language dictionary version control (asset, independent of package version)
 __scripts_version__ = (
     "v0.0.4"  # Git Hook scripts version control (independent from __lang_version__)
 )